SpaceX may be extraordinarily expensive. It may also be building one of the most powerful commercial flywheels we have ever studied.

Those two statements are not mutually exclusive. In fact, the tension between them is exactly what makes the company interesting.

In What Tomorrow Could Become, we promised to reveal the business behind September’s new Tier Two investment thesis.

It is SpaceX...

Paid subscribers received our research on 18 September. Ten days later, Starship supplied an important new piece of evidence: it reached orbit for the first time and deployed 26 Starlink V3 satellites. An engine problem brought the mission to an early end, so the engineering challenge is far from finished. But Starship had moved one step closer from experimental vehicle towards something economically more consequential: a system capable of deploying infrastructure for SpaceX’s own commercial network.

That matters because our thesis has never rested on rockets alone.

Falcon helped build Starlink. Starlink created recurring demand for Falcon. Reuse improved launch economics and generated operating experience that fed back into the system. Starship is an attempt to push that mechanism much further: dramatically more capacity per launch, greater reuse and a lower cost of building whatever comes next.

And then comes the price...

Many investors look at SpaceX’s valuation and stop there. That is understandable. A wonderful company can still be a terrible investment if too much of the future is already embedded in the shares.

SpaceX makes that problem especially demanding because it is the quintessential long-duration equity. Much of what investors are paying for lies years ahead, precisely when rising long-term yields should make distant cash flows less valuable.

But long-duration equities are not bonds.

Their future cash flows can change.

A new capability can create a market that barely existed before. A lower launch cost can make another business viable. A network can generate demand for the infrastructure that expands the network. And an advantage built in one part of the organisation can become the starting point for the next. That is where SpaceX connects directly with our Abundance Shock work.

The investment case therefore turns on a harder question than whether Starship flies or whether another futuristic market eventually appears.

Can SpaceX keep converting difficult-to-replicate capabilities into businesses that make the next capability cheaper, more useful and more valuable?

And if it can, how much of that value ultimately survives for each shareholder after the enormous reinvestment, replacement costs and dilution required to build it?

That is the question the original research below tries to answer. The future may be enormous. SpaceX still has to earn the valuation investors have already placed on it.

Good reading.

Some companies become large. A very small number become powerful enough that states have to reckon with them.

The Dutch VOC and the British East India Company belonged to that rarer category. At their height, they were not merely merchants moving cargo between continents. They fielded armies and navies, negotiated with governments, made war and peace, collected taxes and administered territory. By 1800, the British East India Company governed roughly one-fifth of the world’s population with a military force larger than England’s.

Contemporary scholars describe these organisations as company-states because the boundary between commercial enterprise and sovereign power had become extraordinarily difficult to draw.

Their dominance rested on privileges and practices that have no legitimate modern equivalent, including chartered monopolies, coercion and colonial rule. That distinction matters.

But the economic mechanism behind their rise still deserves attention.

Control access to a difficult frontier and you may eventually influence much of the commerce that develops beyond it.

That is precisely the parallel explored by Stefano Marcuzzi and Alessio Terzi in their June 2026 Cambridge working paper, Outsourcing the Final Frontier. The authors argue that today’s space economy is beginning to display some of the institutional dynamics of the Age of Sail: concentrated infrastructure, dependence on a dominant provider and an increasingly blurred boundary between private capability and strategic national interest. Cambridge summarised the comparison starkly: SpaceX’s grip on a critical transport technology is “almost without precedent” in the past four centuries.

Their concern is primarily about governance.
Ours is about economics.

What happens when the company that dominates access to a frontier can also build the infrastructure, networks and businesses that make that frontier valuable?

That question, of course, brings us to SpaceX. ( )

The company may ultimately prove valuable not simply because it can reach space more efficiently than its competitors, but because access itself can become the foundation from which entirely new businesses are built.

Its three reporting segments, Space, Connectivity and AI, describe different revenue streams. Our investment hypothesis concerns what connects them: the ability to turn one difficult-to-replicate capability into several mutually reinforcing businesses.

Launch supports satellite deployment. Satellite infrastructure supports connectivity.

The company is now attempting to extend those capabilities into computing and intelligence.

The opportunity is substantial. So is the difference between building the infrastructure of the future and earning an attractive return on the capital required.

The Advantage That Builds the Next One

A commanding market share is an outcome. The investment question is whether the forces behind it are strengthening.

Google accounted for 91.1% of worldwide search-engine referrals in Statcounter’s August 2026 data. That measures traffic referred to websites in its sample, not every search performed, but it illustrates the concentration of an important activity. A challenger needs more than a recognisable search box to displace the system surrounding it.

Meta offers the same lesson at a scale that is easier to visualise. During June, an estimated 3.60 billion people used at least one of Facebook, Instagram, Messenger or WhatsApp every day. Against a 2026 world population of roughly 8.3 billion, that is equivalent to about 43% of humanity, or roughly one person in every 2.3 alive. Meta then converted that reach into $59.4 billion of advertising revenue in a single quarter.

A competitor can reproduce a feed. It can copy Stories, short-form video, messaging or almost any individual feature. What it cannot reproduce overnight is the network of users, creators, advertisers, relationships and distribution already assembled around them.

That is when scale stops being merely an output and starts becoming part of the moat.

The businesses are very different, but the underlying mechanism is similar:

Can dominance in access to a market become the foundation of an even larger commercial system?

For SpaceX, that system begins with access to orbit.

Its roadshow reports that its share of global payload mass increased from 45% in 2021 to 65% in 2022 and more than 80% across 2023–2025. This includes its own satellites and development cargo, not just third-party payloads. It is therefore not an equivalent share of commercial launch revenue.

That qualification also reveals part of the advantage.

SpaceX builds a large part of its own launch demand. During the first half of 2026, 61 of its 78 launches were internal, principally supporting Starlink. The launch system helps build the network; the network supplies recurring reasons to use the launch system.

Reusability strengthens the relationship. Recovering expensive hardware allows additional missions without manufacturing an entirely new booster. More flights produce operating experience; inspections feed improvements into the next flight. The IPO roadshow presentation illustrates the result through 165 Falcon-family missions in 2025 alongside eight new boosters flown.

That does not mean eight boosters performed every mission. It demonstrates the importance of the existing reusable fleet.

A rival must therefore assemble more than a rocket. Manufacturing, launch infrastructure, refurbishment, operating experience and downstream demand all matter.

The advantage is not guaranteed to widen every quarter. Second-quarter launches declined year on year (yoy) even as Space revenue increased through customer mix. Our thesis concerns the development of the system, not an uninterrupted rise in every operating statistic.

Starship could change that system’s economics considerably.

Falcon demonstrated the value of recovering and reflying the booster. Starship aims to extend that principle to the complete launch system: the Super Heavy booster beneath the Starship spacecraft, both designed for recovery and reuse. Returning the orbital ship, protecting it during atmospheric entry and preparing it for another mission are central to the programme, not incidental engineering details.

The difference is between spreading some hardware costs across repeated flights and attempting to spread the cost of the whole vehicle across them.

Size adds another dimension. The roadshow compares approximately 23 tonnes to low Earth orbit for an expendable Falcon 9 with an expected 100 tonnes for reusable Starship V3. These are theoretical capacities under specified conditions, not identical operating configurations or achieved Starship commercial performance.

Greater payload capacity and rapid reuse could work together: more infrastructure carried per departure, with less new launch hardware required for each one. Orbital propellant transfer would extend the architecture further, enabling missions beyond Earth orbit. Each capability must still become reliable and economical in practice.

This is why the most revealing comparison concerns the network Starship could deploy. The roadshow shows approximately 2,600 Gbps of satellite bandwidth per Falcon 9 launch, using V2 satellites, against an expected 61,000 Gbps per Starship launch using the developmental V3 configuration. That is more than twenty times the deployed bandwidth.

Installed bandwidth does not invoice a customer by itself. But cheaper, faster deployment of useful capacity could transform the economics of the businesses using it.

An infrastructure asset can become more valuable because it makes other valuable assets cheaper to build.

Some of that value would appear outside the Space segment. Internal satellite launches generate no intersegment revenue; their commercial benefit emerges through the network they help construct.

We should recognise the contribution without inventing internal sales and counting them twice.

Three Businesses, One Expanding System

The financial statements already show why viewing SpaceX solely as a launch provider misses much of the investment.

Second-quarter revenue reached $7.814 billion, up approximately 92% yoy. Connectivity contributed $4.291 billion, AI $2.561 billion and Space $962 million.

Space recorded a $542 million operating loss, alongside $1.076 billion of research and development expenditure as Starship investment accelerated. Its current earnings therefore reflect the cost of developing capabilities intended to support more than external launches.

Connectivity supplies the strongest evidence that the strategy can produce attractive operating economics.

Revenue increased approximately 66%, while operating income reached $1.656 billion, implying a 38.6% operating margin. That is profitability after reported depreciation and share-based compensation, not merely a favourable adjusted measure.

The customer base is also becoming more interesting.

Starlink ended June with 12 million subscriber service lines, twice the number a year earlier. Enterprise & Government revenue grew approximately 108% to $1.806 billion, representing around 42% of Connectivity sales. These calculations show that the business is becoming much more than household internet delivered by satellite.

Aircraft, ships, remote facilities and government operations require connectivity where conventional infrastructure can fall short. A common network can serve customers with very different requirements and willingness to pay. The company’s disclosed applications and recent airline partnerships illustrate that breadth.

However, expansion does not eliminate pricing pressure. Monthly subscriber ARPU was $66, stable sequentially but below $85 a year earlier. Management links geographic expansion to a changing customer mix and local pricing. Subscriber growth must therefore be judged alongside revenue, margins and the investment required to serve each market.

Mobile offers another extension, but management has not disclosed the complete capital requirements of its proposed terrestrial build-out.

Sharing infrastructure can improve economics without making expansion costless.

AI is growing faster, but its returns on capital are less established.

Quarterly revenue increased approximately 247% to $2.561 billion. Adjusted EBITDA turned positive at $1.146 billion, while the segment still reported an operating loss of $1.257 billion and $15.828 billion of capital expenditure. The commercial acceleration is significant. So is the investment burden.

The business pursues two opportunities: selling compute and selling intelligence.

Anthropic and Google are named compute customers in the roadshow, with Google’s disclosed arrangement scheduled to begin in October. Alongside that infrastructure activity sit the group’s own models and applications. SpaceX can earn revenue from other AI developers without requiring Grok to win every market, but it also bears the costs of competing in the model business itself.

The current group has been assembled through acquisitions as well as internal development. SpaceX acquired xAI in February and completed the acquisition of Cursor on 14 August. Integrating these capabilities is an opportunity to demonstrate better economics, not proof that the benefits have already arrived.

Cloud agreements drove most of the latest revenue acceleration. The release identifies $1.6 billion of incremental infrastructure revenue, accounting for most of the AI segment’s year-on-year increase. Its $14.1 billion contracted-sales disclosure includes already recognised revenue and covers enforceable, non-cancellable periods; it is not all additional future revenue.

The more ambitious proposition is to change where that computation takes place.

SpaceX is attempting to turn an energy constraint into a launch opportunity.

Its roadshow argues that orbital computing could benefit from lower-cost solar power, radiative cooling and data distribution through Starlink, provided launch and satellite-production costs fall sufficiently. These are management’s prospective economics, not results from an operating orbital data centre.

The energy logic is important. In suitable orbits, solar panels can receive near-continuous sunlight, avoiding weather and much of the day–night interruption experienced on Earth.

Google’s separate Project Suncatcher research examines precisely that configuration, including a reduced requirement for batteries. That could improve utilisation of solar equipment and reduce dependence on terrestrial electricity infrastructure.

Cooling offers a different potential advantage, but requires a precise explanation.

There is no surrounding air in space to carry heat away. Heat generated by processors must be transported to radiators and emitted as infrared radiation. Properly designed surfaces can reject heat towards space while limiting heat absorbed from sunlight. The process does not require evaporating water into the atmosphere, but it still requires thermal hardware, adequate radiator area and careful engineering.

SpaceX’s proposed design explicitly adds more solar generation and a larger radiator to capabilities retained from its satellite platform. The potential saving is therefore not “free cooling because space is cold”. It is a different power-and-heat-rejection architecture that management believes could lower operating costs.

The complete system must still beat the alternatives after launch, deployment, communications, maintenance and equipment replacement. Google’s research itself identifies unresolved challenges, including high-bandwidth links, radiation and thermal management. Interest from another major technology company strengthens the relevance of the question; it also reminds us that sunlight is not an exclusive advantage.

SpaceX’s potential moat is the ability to assemble the system around it.

The company proposes retaining satellite propulsion, optical links and flight systems while adding compute and larger power-and-cooling equipment. In June, its AI satellite remained at the design stage; in August, Musk expected initial Starmind launches in 2027.

That is the optionality we find valuable: an advantaged starting position for the next business.

Scottish Mortgage’s managers have articulated a related investment interpretation. Tom Slater has emphasised the widening opportunities created by cheaper access to space. Lawrence Burns has highlighted the range of possible businesses rather than dependence on one technological outcome. Their perspective helps frame the question, without settling the answer.

Management’s market estimates illustrate the ambition: approximately $5.7 trillion across its nearer-term opportunities, expanding to $28.5 trillion when enterprise applications are included. The larger figure incorporates the smaller one. These are addressable-market estimates, not obtainable revenue. Beyond them, the roadshow contemplates orbital manufacturing, lunar industry and asteroid mining, whose timing and economics remain far less established.

We do not need to place a discounted-cash-flow value on an asteroid to recognise the potential importance of access.

If such industries develop, SpaceX could participate through transport, communications or infrastructure without becoming the successful operator of every downstream business.

The counterweight is that several opportunities share the same dependencies. Starship success could strengthen them together; persistent delays could weaken them together. They are not independent lottery tickets.

Nor does growing revenue ensure growing shareholder cash flow. Group capex reached $18.369 billion in the quarter. Adjusted EBITDA of $3.538 billion became a $143 million operating loss after depreciation, share-based compensation and restructuring charges. The next phase must demonstrate that additional capital creates more value than it consumes.

What Is the “Right to Build Next” Worth?

SpaceX is difficult to value because we must estimate not only the growth of existing businesses, but how their capabilities could change one another’s economics.

A valuation confined to current profits could miss much of the opportunity.

Capitalising every ambition could miss much of the risk.

The answer is not to abandon valuation. It is to identify the judgements doing the work.

Connectivity provides an operating anchor. Space combines essential infrastructure with substantial development expenditure. AI offers rapidly expanding revenue but much less established cash economics. They should not receive identical valuation multiples simply because they share a parent.

Management’s ambitions provide context, not our base case.

In August, the CFO described a path to a $100 billion annualised revenue run rate by December, including Cursor. That means expected December revenue multiplied by twelve, not annual recurring revenue or $100 billion of recognised 2026 sales.

Musk separately said internal projections for $1 trillion of annual revenue had moved forward to 2030. The public disclosures do not provide enough detail to reconstruct a complete numerical bridge to either outcome.

We can nevertheless reverse the question.

Consider the roadshow’s $135 offering-price reference and the 13.176 billion Class A and Class B shares reported at 30 June. Together, they imply approximately $1.78 trillion of equity value. This is a historical illustration, not the publication-date or fully diluted valuation.

Assume a 12% annual return over five years, no dividends and no subsequent dilution. Equity value would need to reach approximately $3.13 trillion. At 30 times earnings, that requires about $104 billion of annual net income. At a 25% net margin, it requires approximately $418 billion of annual revenue.

These are VMF Research assumptions and calculations, not forecasts. A lower eventual multiple, weaker margins or more shares would increase the required business outcome.

Cursor demonstrates why the ownership denominator matters. The completed transaction was principally share-financed at a $60 billion implied equity value, rather than a $60 billion cash payment. It expanded the shareholder base and brought additional equity awards, so June’s share count cannot simply be carried into September’s valuation unchanged.

This approach makes uncertainty visible without pretending that we know the company’s ultimate value to the nearest dollar.

The premium we are prepared to consider rests on established economics, difficult-to-replicate capabilities and credible opportunities to deploy them elsewhere. It does not rest on the difficulty of the calculation.

Likewise, unmodelled optionality is not automatically free. If the purchase price already requires several unproven businesses to succeed, we are paying for those outcomes whether or not the spreadsheet names them.

Capital allocation will determine how much survives for shareholders.

At 30 June, SpaceX held approximately $100 billion of cash and marketable securities, alongside $39.4 billion of debt and finance leases. That provides considerable execution capacity, but management indicated that each of the following two quarters could involve capex similar to the second quarter.

Liquidity is funding for the strategy, not an uncommitted surplus.

The CFO’s claim of a sub-one-year payback on new compute deployments is encouraging, but the public disclosures do not define the calculation well enough for us to reproduce it. We need the cash evidence.

The adverse outcome need not involve technological failure. SpaceX could remain strategically important while investment, replacement costs and dilution prevent shareholders from receiving the returns they expected.

But demanding that every uncertainty disappear before investing carries its own cost.

Facebook’s early stock-market history illustrates why...

Its shares were offered at $38 in May 2012 and subsequently traded as low as $17.55, approximately 54% below the offering price. By 16 September 2026, the shares stood at $673.31, approximately 38.4 times their post-IPO low, equivalent to roughly 30% annualised price appreciation over fourteen years, excluding dividends.

Knowing the ending makes the decline look easier to endure than it was.

What investors could observe was new operating evidence. By the second quarter of 2013, mobile represented approximately 41% of advertising revenue, while total revenue grew 53% year on year. By the fourth quarter of 2014, mobile’s advertising share had reached approximately 69%. The company was demonstrating how its changing audience behaviour could become a much larger commercial opportunity.

The useful lesson is not that celebrated IPOs recover. Many disappoint permanently. It is that investors can distinguish uncertainty that is gradually being resolved from a thesis that is deteriorating. Facebook supplies a case study, not a forecast for SpaceX.

Our tests are different. Starship must become a repeatable deployment system.

Connectivity must convert added capacity into profitable customer activity. AI contracts must produce attractive cash economics after operating and renewing the infrastructure. Across the group, progress must become visible in per-share value.

Those conditions allow patience to remain a discipline rather than an excuse.

We would reconsider the thesis if deployment delays became persistent, customer economics weakened, or investment and dilution repeatedly outpaced the growth of earning power. A lower quotation would not automatically justify buying more. A higher quotation would not automatically validate the business case.

The allocation reflects conviction in the competitive system, tempered by the unusually wide distribution of outcomes. We are not treating every segment as an established compounder. Connectivity supplies demonstrated profitability; the wider group offers the possibility of applying difficult-to-replicate capabilities across much larger markets. The position must earn additional capital through evidence.

The East India Company comparison brought us here because access can become the foundation of a much larger commercial system. Our thesis does not require SpaceX to own every resource or operate every business beyond Earth. It requires the company to keep turning capabilities that competitors struggle to reproduce into services customers will pay for, at returns that reward the capital committed.

That is the connection with our Abundance Shock thesis: a company that could help expand what becomes economically possible while retaining a valuable position in the infrastructure that makes it happen.

The opportunity is not simply to own a company that reaches the frontier. It is to own a company that can keep moving it.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces “The Company That Builds the Frontier” from the September 2026 issue of VMF’s Security Selection. The underlying research was completed on 18 September 2026 at 4:00 p.m. Eastern Daylight Time and first disseminated to paid subscribers on 18 September 2026 at 9:00 p.m. Eastern Daylight Time. Introductory commentary on subsequent Starship and market developments was added for republication and should be distinguished from the original analysis. Charts, prices and third-party information retain their individually stated reference dates.

The publication contains an investment recommendation concerning SpaceX. It is not personalised investment advice and does not consider any reader’s objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The original recommendation introduced a 2% initial allocation to the Quality Model Portfolio at a published reference price of US$152.71. This is an illustrative research allocation, not a suggested weight for every reader. VMF Research’s Model Portfolios do not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines thematic, fundamental, technological, financial and valuation research within a medium- to long-term framework. It examines the economics of SpaceX’s operating businesses, their potential interaction and the capital required to develop them. Management’s revenue ambitions, addressable-market estimates and prospective orbital-computing economics are not VMF Research forecasts. Sources, material assumptions and limitations are identified in the analysis. Views are reviewed through monthly publications, with weekly or ad hoc updates where material developments warrant reassessment.

The reverse valuation is an illustrative exercise, not a price target or return forecast. Its $135 offering-price reference and June share count are historical inputs, not the September recommendation price or a current fully diluted valuation. The assumed 12% annual return over five years is an analytical hurdle, not an expected or promised outcome. Different margins, valuation multiples, capital requirements or dilution could materially change the results. Historical comparisons, including Facebook’s post-IPO experience, do not predict SpaceX’s returns.

SpaceX involves substantial risks, including launch failures and deployment delays, unproven reuse or computing economics, intense competition, regulatory restrictions, high capital expenditure, infrastructure replacement costs, acquisition integration, financing needs, dilution and valuation compression. Its businesses share important dependencies, so their risks are not independent. As a long-duration equity, its valuation is particularly sensitive to required returns and the timing and scale of future cash flows. Technical progress, successful launches or revenue growth may not translate into attractive shareholder returns. Investors may lose some or all of their capital.

Disclosure of interests: as of the original research cut-off, legal entities controlled by Vasco Marques de Freitas held long positions in Scottish Mortgage Investment Trust PLC and Ark Innovation ETF, which both hold SpaceX and therefore provide those entities with indirect economic exposure to the company. This constitutes a financial interest and potential conflict of interest that readers should consider when evaluating the recommendation. Ownership of the Trust and the ETF is not equivalent to direct ownership of SpaceX. The interest does not validate the analysis, reduce investment risk or imply suitability for any reader.

Neither VMF Research nor the author received compensation from any issuer covered in connection with the preparation of the original research, and no issuer reviewed, approved or amended its investment conclusions before first dissemination.

Past performance, Model Portfolio performance and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

Imagine being right about AI, right about economic growth, and still disappointed by the companies you bought.

Cheaper intelligence could open entire markets. It could also hand the savings to customers, invite new competition or leave shareholders financing an expansion whose rewards keep receding. September’s VMF’s Security Selection, “Who Owns Abundance?”, begins with that ownership problem.

We stopped September’s public Tier One series before sharing more of the conceptual work behind our main hypotheses about money, economic growth and market behaviour. That developing framework, and its portfolio implications, remain with subscribers.

For Tier Two, we are taking a different route: we will reveal the company behind our latest investment thesis in the next article. But first, we need to explain the economic context that led us to it.

Anthropic’s scenario work gives this investigation an arresting starting point: an extreme case in which annual US growth eventually reaches approximately 15%. This is a scenario, not a forecast, and the authors assign no probabilities to the outcomes they explore. Our investment case does not require that number to be right.

What interests us is how businesses might respond when activities that were prohibitively expensive become affordable. The consequences could reach well beyond doing today’s work a little faster.

The excerpt below was first published for paid subscribers on 18 September 2026. It sets up the question we must answer before introducing the company:

When cheaper intelligence makes more of the economy possible, who gets to own the parts it still cannot reproduce?

Good reading.

The future is usually discounted in basis points.

Anthropic is asking us to consider it in trillions. In the September issue of VMF’s Strategic Asset Allocation, we approached the problem from the perspective of valuation.

A fixed-rate bond cannot decide that the world has become more productive and increase its coupon. On the other hand, a company can develop a better product, discover a cheaper way to operate or enter a market that barely existed when the original investment was made. Higher interest rates make distant cash flows less valuable, all else equal. But everything else does not have to remain equal.

Future profits can become larger, arrive sooner or become considerably more probable.

Anthropic has now published an economic exercise that allows us to ask a more ambitious question.

What could tomorrow actually become?

The company’s new Economic Scenario Explorer models how advances in artificial intelligence could affect the US economy through 2030. It begins not with a forecast for the S&P 500 or another estimate of how many billions will be spent on data centres, but with something more fundamental: the work an economy actually performs.

Jobs are collections of tasks. Some tasks can be augmented by AI. Others can be automated. New ones can emerge. Productivity changes as people and machines divide that work differently. Adoption determines how much of the available technology actually reaches the economy.

Change those assumptions and the range of possible outcomes becomes enormous.

Anthropic highlights three.

In its modest scenario, AI produces gains broadly comparable with those of an important previous technological revolution. US GDP reaches approximately $34.1 trillion in 2030, only 1.6% above the level generated by the same economy without AI.

The substantial scenario moves much further. AI becomes capable of performing a large proportion of knowledge work, although adoption remains incomplete. GDP reaches approximately $36.3 trillion, 8.3% above the no-AI baseline.

Then comes the extreme scenario.

Here, AI becomes more productive than humans across the great majority of knowledge work, operates with considerable autonomy and is adopted rapidly. US GDP reaches approximately $44.4 trillion in 2030, 32.4% above the counterfactual without AI. As adoption spreads, annual GDP growth eventually reaches approximately 15%, fast enough for the economy to double in size roughly every four and a half years.

Those numbers deserve two reactions.

The first is obvious. They are extraordinary.

The second is more important. They are not forecasts.

Anthropic explicitly assigns no probabilities to the scenarios. Its model is a deliberately simplified representation of a complicated economy and excludes forces that could materially alter the outcome, including policy responses, business cycles, aggregate-demand effects and financial disruptions. The most extreme case also relies on assumptions about recursive AI improvement and rapid adoption that may never materialise. Some of the economists who reviewed the framework considered that scenario closer to a thought experiment than a conventional forecast.

We should therefore resist the temptation to take the most exciting number and build an investment thesis around it.

We do not need to. What matters is that the range of economically serious outcomes may be widening.

Cathie Wood has been making a related argument from a very different starting point. ARK’s framework focuses on the convergence of artificial intelligence, robotics, energy storage, blockchain technology and other innovation platforms. In her latest In The Know, Wood notes that global real GDP growth has averaged roughly 3% over the past 125 years and argues that the rate could at least double. She goes further still, describing the 10% to 15% growth rates discussed by Elon Musk as within the conceivable range.

We made the necessary qualification in Tier One and will repeat it here.

Our investment case does not require any of those numbers to be right.

An economy growing at 4% rather than 3% can still produce extraordinary companies. A business can compound revenues at 20% while the wider economy grows at 2%. New markets, changes in market share and declining costs can produce exceptional corporate outcomes without anything resembling Anthropic’s extreme scenario.

The relevance of these possibilities lies somewhere else.

Conventional economic analysis spends considerable time worrying about the left tail of the distribution. Recession. Financial crisis. Inflation. Debt distress. Geopolitical shock.

Those risks remain.

But technological progress is forcing investors to take the right tail more seriously too.

What if tomorrow is not merely a little more productive than today?

What if entire categories of activity become cheap enough to perform for the first time?

What if falling costs do not simply allow companies to do the same work with fewer resources, but make previously uneconomic work worth doing?

That is the possibility our Abundance Shock has been designed to investigate.

In Tier One, we described the mechanism in deliberately simple terms. As useful capabilities become dramatically cheaper, projects that previously failed an investment committee’s financial test can become viable. Smaller organisations gain access to tools that once required large teams and substantial amounts of capital. Existing businesses can attempt things that would previously have consumed too much money or too much time. The first effect is lower cost. The more powerful effect begins when behaviour changes because the lower cost makes more economic activity possible.

Anthropic’s scenarios take that logic and ask what happens when it spreads across an entire economy.

For investors, however, this is where the easy part ends.

A Bigger Pie Does Not Mean Equal Slices

A much richer economy does not tell us who becomes richer.

Anthropic’s model makes that distinction unusually visible. Today, approximately 60 cents of every dollar produced in the US economy goes to labour and 40 cents to capital. In Anthropic’s modest scenario, that relationship barely moves. Labour receives 59.4% of GDP and capital 40.6%.

The substantial scenario begins to look different. Labour’s share falls to 56.1%, while capital rises to 43.9%.

Under the extreme scenario, the change is profound. The economy is far larger, but only 45.2% of GDP accrues to labour. Capital receives 54.8%. Its share of the economic pie has increased by almost fifteen percentage points.

The social implications are substantial. Anthropic’s extreme scenario combines much greater aggregate prosperity with weaker outcomes for many knowledge workers. It is precisely why the company emphasises how the benefits of technological progress might eventually be distributed.

Our question in VMF’s Security Selection is narrower.

It is an ownership question.

And we need to be careful with the answer.

A larger share of income going to capital does not mean every owner of capital wins.

Far from it.

A technology revolution can create enormous economic surplus while destroying the economics of the companies that originally appeared best positioned to capture it. Competition can force productivity gains through to customers. New entrants can use lower costs to attack incumbent margins. Capital-intensive businesses can spend so aggressively to preserve leadership that shareholders wait years for the promised cash flows to arrive.

The economy can become vastly more productive while individual investments disappoint.

That distinction runs directly through the argument we developed in last month’s issue, “Quality Has a Half-Life”.

August asked a deliberately difficult question:

Would the company become more or less relevant if intelligence became ten times cheaper?

We were trying to identify the scarce assets, proprietary information, trusted relationships and essential processes whose economic importance could increase as intelligence itself became abundant.

September asks the complementary question.

If intelligence becomes ten times more productive, which scarce assets become more valuable?

That is where the distribution of future profits will be decided.

Some of the additional value may accrue to the companies producing intelligence itself. Some may accrue to semiconductor manufacturers and the owners of the equipment required to produce those chips. Some may flow towards electricity, transmission networks and the physical infrastructure supporting an increasingly computational economy.

Other businesses may capture value through intellectual property, proprietary data, distribution, customer relationships, network effects or regulatory permissions. And some will possess combinations of those advantages that reinforce one another.

The result will not be an even distribution of abundance.

It will be a competition to own the bottlenecks it creates.

Growth Is Not the Investment Thesis

This is where Anthropic, ARK and VMF Research arrive at related questions from different directions.

Anthropic begins with tasks. How capable does AI become? How autonomously can it work? How rapidly is it adopted? How much productivity does it create?

ARK begins with technological convergence. Artificial intelligence becomes more powerful as the cost of computation falls and interacts with advances elsewhere in the economy. Wood’s conclusion is that the combined productivity effect could move growth materially beyond the rate investors have come to regard as normal.

Our Abundance Shock begins with economics...

What changes when a useful capability becomes dramatically cheaper?

The three frameworks do not need to produce the same forecast to point towards the same possibility.

We may be approaching a period in which the productive frontier moves faster than conventional economic models, corporate planning cycles and financial markets are accustomed to handling.

But growth itself is not the investment thesis.

Between greater economic output and greater shareholder wealth sits a sequence of increasingly difficult tests.

That last step is the one an investor cannot skip.

A wonderful technology does not guarantee a wonderful business. A wonderful business does not guarantee wonderful shareholder returns. And an economy capable of growing much faster than we expected does not remove the requirement to determine what we are paying for the claim on that future.

There is another implication, and it may prove even more important.

Abundance can make scarcity more valuable.

Cheaper scientific discovery can create more compounds worth testing, increasing demand for clinical trials, manufacturing capacity and regulatory infrastructure. Cheaper engineering can produce more viable physical projects, increasing demand for electricity, materials, construction and logistical capacity.

Cheaper intelligence can allow businesses to perform far more work, while increasing the value of the networks, permissions, physical assets and infrastructure through which that work must ultimately pass.

We saw an early version of this relationship in August. Veeva did not need to own the foundation model. Its opportunity came from controlling regulated workflows through which increasingly capable models must pass before they can create economic value in life sciences.

Generic intelligence may become cheaper. Completed, compliant work remains scarce.

The same principle can operate at much greater scale.

If the range of economically viable activity expands, the most valuable businesses may not simply be those creating the abundance. They may be the ones controlling what abundance cannot easily reproduce.

• Physical infrastructure.

• Distribution.

• Networks.

• Manufacturing capability.

• Proprietary technology.

And, in the rarest cases, several of those advantages reinforcing one another inside the same organisation. That observation led us towards the company at the centre of this month’s recommendation.

Its ambition is enormous.

So is the uncertainty surrounding what it may ultimately become.

Our task is not to decide whether the future it imagines sounds exciting. It is to determine whether the advantages it is building can allow shareholders to capture enough of that future to justify the price they are being asked to pay today.

The future may become abundant.

The businesses capable of owning its bottlenecks will not be.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces “What Tomorrow Could Become” from the September 2026 issue of VMF’s Security Selection. The underlying research was completed on 18 September 2026 at 4:00 p.m. Eastern Daylight Time and first disseminated to paid subscribers on 18 September 2026 at 9:00 p.m. Eastern Daylight Time. Charts, economic estimates and third-party research retain their individually stated reference dates.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. VMF Research’s Model Portfolios are illustrative research portfolios, not client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines macroeconomic, technological, fundamental and scenario-based research within a medium- to long-term framework. The Abundance Shock is a working hypothesis, not an assured economic outcome. Views are reviewed through monthly publications, with weekly or ad hoc updates where material developments warrant reassessment. References to Anthropic, ARK, Cathie Wood and other third parties do not imply their endorsement of, or participation in, VMF Research’s conclusions.

Anthropic’s economic scenarios are not forecasts and carry no assigned probabilities. The approximately 15% annual US growth rate belongs to its extreme scenario, which depends on demanding assumptions about AI capability, autonomy and adoption. Estimates of additional output in 2030 are comparisons with a modelled counterfactual, not annual growth rates. Anthropic’s scenarios, ARK’s projections and VMF Research’s analysis use different assumptions and should not be treated as interchangeable or as validation of one another.

Greater productivity or a larger share of income accruing to capital does not guarantee higher shareholder returns. Competition, implementation costs, capital expenditure, technological disruption, regulation, financing needs, dilution and excessive valuations can prevent investors from capturing the benefits of economic growth. Outcomes may differ materially from the scenarios discussed, and investments may lose some or all of their value.

Disclosure of interests: legal entities controlled by Vasco Marques de Freitas held a long position in Scottish Mortgage Investment Trust PLC at the 18 September research cut-off. The Trust provides indirect exposure to Anthropic, whose economic research is discussed in this article. Those entities also disclosed a long position in ARK Innovation ETF (ARKK) in the 11 September Tier One publication; this article discusses ARK’s economic framework. These interests may create potential conflicts and should be considered when evaluating the analysis. They do not validate its conclusions or imply suitability for any reader.

Neither VMF Research nor the author received compensation from any issuer covered in connection with the preparation of the original research, and no issuer reviewed, approved or amended its investment conclusions before first dissemination.

Past performance, Model Portfolio performance and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

A thirty-year Treasury is not necessarily held by someone with thirty years to spare.

Some hedge funds finance those bonds with short-term borrowing. Tighter funding or larger collateral demands can turn a willing lender to Washington into a forced seller. For these investors, volatility can determine whether they are able to keep holding the debt at all.

This complicates the contrast we explored in A Hawk on Borrowed Time. Kevin Warsh wants clearer market signals, less shaped by Federal Reserve guidance. Scott Bessent must keep refinancing the government in the market producing those signals.

The Treasury describes its expanded buybacks as liquidity support. Our research asks how far that support can stabilise the long end without obscuring the information in its prices. The funding deserves equal attention: issuing new bills and drawing down cash held at the Fed have different consequences for the surrounding financial system.

So... can Washington obtain more dependable financing without making its creditors more dependent on intervention?

First published for paid Tier One subscribers on 11 September 2026, this excerpt follows that question beyond buybacks, into changing reserve preferences and the digital dollars bringing new buyers to American debt. It advances the investigation behind our Dovish Shock thesis. The full September issue carries the analysis through to the Model Portfolio.

Good reading.

The Treasury does not get to wait for the market to agree with it. America’s bills fall due anyway, and every refinancing forces Scott Bessent to confront the price investors demand to keep lending.

What Warsh would like to recover as an economic signal arrives at the Treasury as a financing cost.

On 19 August, Bessent’s department responded. The Treasury announced that the maximum size of its long-end liquidity-support buybacks would increase from $2 billion to at least $4 billion per operation, covering the ten-to-thirty-year sectors from 9 September through 4 November.

Its stated purpose was to improve liquidity.

For our Alpha Tier subscribers, the significance extended beyond those purchases. We interpreted the announcement as evidence that the administration’s tolerance for rising long-term yields had limits. Investors watching the Fed for signs of accommodation had received an answer from the other side of Washington.

The financing mechanics deserve precision.

To understand what these buybacks can achieve, follow the cash. The Treasury pays investors to take bonds off their hands. If it funds those purchases by selling short-term bills, it replaces longer-term debt with shorter-term debt.

The Treasury can also draw on cash already held in its account at the Fed, known as the TGA. Drawing down that balance releases funds into the banking system. This can ease financial conditions without a Fed rate cut. But borrowing to replenish the account can absorb that liquidity again.

Buybacks can provide relief... they do not solve the budget deficit.

On 24 August, five days after the announcement, Stanley Druckenmiller challenged the plan in The Wall Street Journal. The criticism was especially pointed. Druckenmiller had worked alongside Bessent at Soros Fund Management and understood the pressures facing his former colleague.

He called the initiative “price management”.

His argument was straightforward: investors were demanding higher yields for reasons Washington needed to address. Inflation and persistent deficits were influencing the price they charged to lend.

Trying to suppress that price risked silencing the warning while leaving its causes intact.

We share the concern about credibility.

Well-designed liquidity support can strengthen a market. The danger begins when investors conclude that the government finds the market’s verdict unacceptable and intends to keep intervening until the price becomes more convenient.

A bondholder might welcome the immediate support while becoming less comfortable with the currency in which that bond will be repaid...

The dollar fell after Bessent’s announcement on 19 August. Its subsequent behaviour brings us back to the puzzle we exposed earlier in the What the Headlines Miss section. The chart on page 6 shows the summer rebound faltering, even as long- term Treasury yields remained elevated inside our TACO Zone.

Higher yields can make dollar assets more attractive. But the extra interest is only part of the return. For investors abroad, losses on the dollar can outweigh that advantage. And higher yields do not always signal confidence in stronger growth. They can also reflect demands for compensation against inflation and the risks of persistent government borrowing.

That is the possibility we are watching. High yields and a weakening dollar can express the same unease. The charts alone cannot establish the cause, but they help explain our concern: supporting Treasury bond prices may offer immediate relief without restoring confidence in the money those bonds promise to repay.

For a central bank, this raises a practical question about how to protect its country’s reserves. These assets form a financial buffer for difficult times. Earning interest is useful, but so is reducing dependence on any one currency or government.

Gold offers that independence. It pays no interest, and its price can fall sharply. Yet physical bullion held outright is nobody else’s debt. There is no issuer who must honour a payment or find new lenders when an old obligation falls due.

Its growing role is already visible. According to the ECB, gold represented 27% of global official reserves at the end of 2025. US Treasuries accounted for 22%. Rising gold prices explain much of that shift: the metal already held became more valuable. Central banks also continued buying additional gold, although purchases slowed from the exceptional pace of previous years. The change therefore reflects both accumulation and appreciation.

China makes the accumulation easier to see.

Its official figures show another 650,000 ounces added during August, taking reported holdings to 76.73 million ounces. Those are additional ounces, so the increase cannot be explained by a rising gold price.

Alongside the decline in China’s reported Treasury holdings shown earlier, this is consistent with a gradual diversification of its reserves. We cannot establish that each dollar withdrawn from Treasuries went into gold. But we can observe a larger holding of physical metal alongside a smaller reported holding of American government debt.

Even the location of the metal is receiving renewed attention.

On 2 September, the Dutch central bank disclosed a redistribution that reduced New York’s share of its gold reserves from 31.3% to 18.5%. Most of the adjustment involved sales in New York and purchases in London, alongside physical transfers. Total gold holdings were unchanged. The bank cited crisis preparedness and the ability to mobilise its reserves quickly.

That is a revealing consideration. An emergency reserve must be accessible when financial conditions are strained and political relationships may be least predictable. Ownership, custody and liquidity all enter the decision.

Bessent, meanwhile, has a very modern way of reaching the next buyer of American debt...

Consider a saver outside the United States who wants to hold dollars on a phone. That person may never open an American brokerage account or participate in a Treasury auction. Yet buying a dollar stablecoin can place their savings behind US government debt through the issuer’s reserve portfolio.

The GENIUS Act establishes a framework requiring at least one-for-one backing with eligible reserve assets, including cash and short-dated Treasuries. It gives the expansion of digital-dollar payments a direct connection to demand at the short end of the government bond market.

Bessent has made the policy objective explicit. When the legislation was signed, he linked stablecoin adoption to both stronger international use of the dollar and increased demand for Treasuries.

Washington’s enthusiasm has a financing logic: a larger digital-dollar economy can bring a larger audience to America’s debt.

The opportunity is substantial, but the net contribution depends on where the money originates. Moving savings from a money-market fund into a stablecoin may simply move Treasury ownership between intermediaries. Treasury’s own advisory committee has acknowledged that offset. And a growing appetite for bills does not automatically supply buyers willing to lend for thirty years.

For our Alternatives book, these tensions create opportunities with different drivers.

Gold draws support from the search for an asset that carries no sovereign promise. Bitcoin offers a digital form of monetary scarcity. And the networks carrying stablecoin payments may benefit as more financial activity moves onto them. Here, however, adoption must translate into lasting economic value for their tokens.

These assets do not need a single forecast to rise together. Some investors are seeking protection from the monetary system’s vulnerabilities. Others are investing in the technology helping that same system extend its reach.

Both forces can operate at once.

That brings us closer to resolving the puzzle we framed in “What the Headlines Miss”. It helps explain the simultaneous strength of precious metals and crypto. Yet another piece remains: the resilience of equities whose valuations depend heavily on profits many years into the future.

Elevated long-term yields should make those distant earnings less valuable today, all else equal. But markets rarely hold everything else equal. The question is whether investors are seeing something in tomorrow’s earnings powerful enough to outweigh today’s higher cost of money.

To understand that, we must look beyond Washington’s efforts to finance yesterday’s commitments. We must examine what the economy may be capable of producing tomorrow.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces “Old Debts, New Dollars” from the September 2026 issue of VMF’s Strategic Asset Allocation. The underlying research was completed on 11 September 2026 at 4:00 p.m. Eastern Daylight Time and first disseminated to paid subscribers on 11 September 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not take account of any reader’s objectives, financial circumstances, knowledge, experience, liquidity needs or risk tolerance. The Tier One Model Portfolio is an illustrative research portfolio and does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines macroeconomic, monetary-policy, fixed-income, liquidity, reserve-management and digital-asset research. Conclusions concerning Treasury buybacks, the Treasury General Account, long-term yields, changes in Treasury demand, reserve diversification, gold, stablecoins and digital assets are analytical judgments and conditional scenarios rather than assurances.

Treasury buybacks are debt-management operations and do not, by themselves, constitute quantitative easing or yield-curve control. Their effect depends on scale, maturity, financing and surrounding issuance. A TGA drawdown can add bank reserves, all else equal, but later borrowing or other Treasury operations can offset that effect. Buybacks may improve liquidity without sustainably lowering long-term yields and may also affect perceptions of fiscal and monetary credibility.

Gold, Bitcoin, cryptoassets, stablecoins and related instruments involve materially different risks, including market, liquidity, custody, counterparty, technological, cybersecurity, regulatory and currency risks. Stablecoin growth does not necessarily create net new Treasury demand, and greater demand for short-dated bills does not necessarily translate into demand for long-dated government bonds.

Disclosure of interests: as of the research cut-off, legal entities controlled by Vasco Marques de Freitas held long positions in listed exchange-traded products providing economic exposure to Bitcoin, Ether and Solana. These interests may create actual, potential or perceived conflicts of interest and should be considered when evaluating the analysis. Their existence does not validate the conclusions, reduce investment risk or imply suitability for any reader.

Neither VMF Research nor the author received compensation from any issuer, fund sponsor or other covered entity in connection with the preparation of the original research, and no covered entity reviewed, approved or amended its investment conclusions before first dissemination.

Past performance, Model Portfolio performance, historical market relationships and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

A rate hike is an awkward headline for a thesis called the Dovish Shock.

On 16 September, the Fed raised rates to 3.75-4.00%, with policymakers projecting another increase this year. The hawkish interpretation has substance.

Five days earlier, paid Tier One subscribers received The Price of Credibility. Its opening takeaways included a possibility worth revisiting now: “Our Dovish Shock could begin with a rate hike!”

We retain that conviction. Our argument concerns how long restraint can endure as maturing debt is refinanced at prevailing rates. Establishing credibility may give Warsh greater freedom to ease later, while mounting financing pressure could make that freedom increasingly necessary. The latest hike makes this a harder test; it does not establish that our scenario will follow.

September’s issue examines the competing demands behind that test: a Fed rebuilding its authority, a Treasury continually seeking buyers and an AI economy consuming capital today while promising greater productive capacity tomorrow.

Following The Treasury Just Blinked, this excerpt asks how much financial strain restoring market discipline might require. The full issue follows the consequences into monetary alternatives, the Abundance Shock and the Model Portfolio implications.

What follows was first published on 11 September 2026, before the rate decision. It explains why we remain focused on the constraints that could eventually change Warsh’s calculation.

Good reading.

Kevin Warsh wants markets to tell him something he has not already told them.

To understand the ambition, imagine a copper mine closing and the price of copper jumping. On another continent, an engineer changes a design, a manufacturer searches for a substitute and a recycler finds that yesterday’s scrap is worth collecting.

They need never learn what happened at the mine...

The price carries the news, and the incentive to respond.

Friedrich Hayek recognised the extraordinary coordination hidden inside these ordinary decisions. Knowledge scattered across millions of people could influence economic activity without first passing through a central authority. This remains one of the Austrian School’s most powerful insights.

Interest rates extend that coordination across time. They influence which investments receive scarce capital, which borrowers must wait and which promises become too expensive to finance. As believers in free markets, we admire Warsh’s attempt to recover more of the information those prices can convey.

At Jackson Hole, he called for “clear market signals, as unfiltered as possible”. His retreat from habitual forward guidance asks investors to form more independent judgements while preserving the Fed’s freedom to respond to an uncertain economy.

Yet investors have learned to examine a jobs report through the anticipated response of the Fed. They trade that expectation. Policymakers then examine the resulting prices for information about the economy.

The Fed can find itself listening to an echo of its own voice.

Ben Bernanke warned about this “hall of mirrors” in 2004. Add the expectation of a Fed put, the belief that sufficiently severe market losses will eventually summon support, and tomorrow’s anticipated intervention starts influencing today’s willingness to take risk and the prices they are prepared to pay today.

A quieter chairman cannot dismantle that expectation through silence alone.

Allowing prices to speak means being prepared to hear an uncomfortable answer.

For Washington, that answer arrives at each debt auction.

Gross federal debt stood at approximately $40.1 trillion on 3 September, including $32.4 trillion held by the public. The carrying cost is already substantial: CBO estimates net interest expenditure of $963 billion in the first ten months of fiscal 2026, equivalent to more than a fifth of federal revenue over that period.

A low fixed coupon can shelter a borrower from changing conditions for years. Then the bond matures. Principal comes due, and replacement borrowing must compete at the prevailing price. Another piece of yesterday’s financing is exposed to today’s market.

Treasury bills bring that negotiation back sooner and more frequently.

Alongside persistent deficits, the refinancing calendar keeps renewing the government’s need for willing lenders. Higher borrowing costs gradually become budget expenditure, competing with the priorities politicians would rather fund.

Warsh can ask for patience while monetary restraint works. The refinancing calendar does not pause.

The creditors arriving at those auctions are also changing.

Over the year to June 2026, reported Treasury holdings in China and Japan declined, as did aggregate foreign official holdings. Total foreign holdings nevertheless rose from approximately $9.09 trillion to $9.30 trillion.

America continues to attract overseas capital but with a greater share of those holdings resting with private investors.

The terms on which that capital participates deserve attention.

For leveraged buyers, demand can depend on the cost and availability of their own financing. Washington’s access to funding can become entangled with its creditors’ access to funding, adding another channel through which pressure can spread.

Against this backdrop, we believe the hawkish interpretation of Warsh risks becoming too static.

A chairman dismissed as a Trump proxy risks having every rate cut interpreted as obedience. A chairman who has demonstrated independence has more room to argue that circumstances have changed. Establishing credibility now can increase his freedom to act later.

His determination to confront inflation can be sincere and still precede a forceful turn towards accommodation.

Our Dovish Shock thesis rests on that possibility.

We believe investors risk overestimating how long restraint can endure as refinancing costs feed through the system. If funding pressure broadens into credit and economic activity, we expect the Fed to turn more decisively than a simple extrapolation of Warsh’s present stance would suggest.

The potential surprise lies in the transition from establishing credibility to using it.

There are limits to this interpretation...

The Fed can support markets without cutting rates. That alone would not confirm the broader easing we expect. The test comes as America replaces maturing debt.

Buyers may remain willing, and borrowing costs may stay manageable. Therefore, if the Fed also keeps rates high through those refinancings, our thesis would weaken.

For now, we retain conviction in our Dovish Shock thesis. We do not dismiss every hawkish statement as theatre or expect every market setback to trigger a Fed rescue.

Our judgement concerns the constraints that may eventually change the monetary policy calculation:

Kevin Warsh is trying to restore authority to the price of money. His credibility will be tested by the consequences of allowing that price to adjust.

Scott Bessent faces those consequences on a more immediate timetable. He must keep finding buyers.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces “A Hawk on Borrowed Time” from the September 2026 issue of VMF’s Strategic Asset Allocation. The underlying research was completed on 11 September 2026 at 4:00 p.m. Eastern Daylight Time and first disseminated to paid subscribers on 11 September 2026 at 9:00 p.m. Eastern Daylight Time. Charts and data retain their individually stated observation dates. Introductory commentary concerning subsequent policy developments was added for republication and should be distinguished from the original research.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. References to the Tier One Model Portfolio concern an illustrative research portfolio, not client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines macroeconomic, monetary-policy, fixed-income and funding-market research within a medium- to long-term framework. Assessments of policy credibility, market signals, government refinancing and investor behaviour reflect VMF Research’s interpretations and conditional hypotheses, rather than official guidance or assurances. Views are reviewed through monthly publications, with weekly or ad hoc updates where material developments warrant reassessment.

Government bonds and other interest-rate-sensitive investments involve market, duration, inflation, liquidity and currency risks. Leveraged exposures introduce additional collateral, refinancing and forced-sale risks, even without deterioration in the underlying issuer’s creditworthiness. Historical relationships between borrowing costs, financial conditions and asset prices may change, and investors may lose some or all of their capital.

Neither VMF Research nor the author received compensation from any issuer, fund sponsor or other entity covered in connection with the preparation of the original research, and no covered entity reviewed, approved or amended its investment conclusions before first dissemination.

Past performance, Model Portfolio performance, historical market relationships and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

The inflation thesis met its counterargument

Begin with When Contrarian Becomes Consensus, from August’s Strategic Asset Allocation issue.

Our original framework treated the pandemic as a break with the previous investment regime. Inflation and interest rates would become higher and more volatile, while physical resources and productive capacity would matter more. That thinking shaped the portfolio’s exposure to alternatives and natural resources, alongside its relatively small conventional fixed income allocation.

By August, those arguments had become considerably less unusual. The question was whether investors had become so accustomed to finding inflation everywhere that they were overlooking what might push in the opposite direction.

Lacy Hunt’s evolving framework made that question particularly interesting. In The Old Deflation Map, we examined how an economist associated with debt driven disinflation was placing greater weight on constraints in supply. Debt could still weaken demand, but its consequences depended on the economy’s ability to expand production.

His analysis also raised a serious objection to the AI optimism developing elsewhere: the spending required to build the technology was arriving before its full productivity dividend.

That tension led into The Hawk and the TACO Zone. Our question was whether the market was interpreting Kevin Warsh’s commitment to price stability too mechanically. A firm inflation target does not prescribe the same interest rate under every economic outcome.

The Dovish Shock remained a hypothesis about expectations. A market anticipating renewed tightening could reprice meaningfully if the eventual policy path proved less restrictive, even without an aggressive cutting cycle.

The potential economic reason for that repricing appears in Part II of The Abundance Shock.

Building AI requires scarce electricity, equipment, construction capacity and capital. Deploying it could allow businesses to complete more work without increasing their inputs proportionately. The same technology can therefore intensify bottlenecks during construction while helping to reduce costs elsewhere.

The important question is how quickly those benefits spread. Cheaper software does not automatically produce a falling general price level, and promising productivity data do not establish that AI caused the improvement. We examined both the opportunity and the reasons it might take longer than enthusiasts expect.

That explains the paid issue’s title, Barbell. We retained exposure to physical scarcity while also owning businesses and sectors that could benefit from cheaper intelligence. The Model Portfolio did not require us to declare one side permanently victorious before the evidence justified it.

Quality has a deadline

Two weeks later, Security Selection encountered a related problem inside company analysis.

Terry Smith had built Fundsmith around buying good businesses, avoiding excessive prices and then allowing compounding to work. Yet the fund reported 51.8% portfolio turnover during the first half of 2026 and announced a greater emphasis on fundamental and share price momentum.

When “Do Nothing” Stops Working examines what changed, including the pressure that redemptions can impose on a manager whose investments need time.

We did not conclude that patience had become obsolete. We concluded that patience requires more demanding evidence.

A recoverable setback needs an intact franchise, capable management and enough time to repair the damage. AI may compress that third condition. A company can retain attractive historical margins while competitors become faster, customers reconsider alternatives and the window for a turnaround narrows.

That gave us a useful test: would this business become more valuable or less valuable if intelligence became dramatically cheaper?

Following it led us back into life sciences, but beyond the biotechnology companies introduced in July. The new recommendation was an established business embedded in the specialised work surrounding the industry, where useful intelligence must operate alongside permissions, traceable evidence and trusted processes.

Its opportunity depends on capturing part of the value AI creates. Helping customers become more productive is not sufficient if competition forces all the benefit into their hands. The paid issue examined that distinction through the business economics, valuation scenarios and risks before making a measured initial allocation. The company’s identity and complete investment case remain reserved for subscribers.

The same discipline went back through existing recommendations. Regeneron and Halozyme supplied stronger operating evidence, while disappointing quarters at Universal Music Group and Amrize required a less comfortable examination of margins, cash conversion and execution. A rising share price deserves scrutiny just as a falling one does.

A gold buyer inside the crypto economy

Alpha Tier approached another familiar investment argument through The End of Simple Frameworks.

Tether provided the case study. A company built inside digital finance was also accumulating physical gold, while Treasury securities remained central to the reserves supporting its dollar business. Its balance sheet made the usual argument over whether gold or Bitcoin must “win” look rather incomplete.

We examined the different jobs those assets can perform, and the equally important question of how they are owned. A scarce asset placed inside a financed company acquires risks that do not belong to the asset itself. Debt, distributions and redemption obligations can force decisions that a patient owner might otherwise avoid.

This connected directly with the Fundsmith discussion. An investment’s prospects cannot be separated entirely from the demands placed on its owner. Understanding the asset is only part of understanding the position.

The Treasury gave us a new signal

The final excerpt is The Treasury Just Blinked.

On 19 August, Treasury announced that liquidity support buybacks in the longer maturity sectors would increase from a maximum of $2 billion to at least $4 billion per operation, beginning in September. Its stated purpose was to provide greater liquidity support in those markets.

Our August Alpha Tier interpreted that announcement through the TACO Zone framework: the area where long term borrowing costs become consequential for mortgages, investment, valuations and government financing. We saw another channel through which financial conditions might change, without waiting for the federal funds rate to move.

The distinction matters. Treasury buybacks are not quantitative easing or a promise to cap yields. Nor does a potential drawdown in the Treasury General Account establish that a particular amount of liquidity will be released.

The issue also examined the contrary case, articulated by Stanley Druckenmiller: intervention intended to improve market functioning could ultimately make investors demand more compensation for fiscal and policy risk. That would have very different implications for bonds from the benign liquidity scenario.

We regarded the announcement as supporting evidence for the Dovish Shock, not its completion. Warsh’s Jackson Hole speech on 28 August maintained the emphasis on inflation and made clear that better summer readings had not convinced him that the underlying trend had improved sufficiently. The test remained live.

What the research changed

August produced a new company recommendation, but it did not produce wholesale Model Portfolio changes.

Tier One retained its balance between scarcity and abundance. Alpha Tier examined improving signals across precious metals, miners and crypto, then left its allocations unchanged. Existing exposure was already meaningful; stronger confirmation alone did not make increasing it the best use of capital.

That is the practical thread connecting the month. We challenged the macro framework, tightened the test for business durability and reconsidered the portfolio implications as policy and prices evolved.

The seven excerpts give you a substantial body of that research to examine for yourself. The complete August issues also contain the new life sciences recommendation, its valuation and risks, the reviews of existing positions and the analysis behind the allocation decisions.

Important Disclosure

This article was produced by VMF Research, Lda., under the responsibility of Vasco Marques de Freitas, CFA, CMT, Founder and CEO. It contains general investment research and commentary, including summaries of investment recommendations. VMF Research’s activity as a producer of investment recommendations is subject to the European market abuse framework, including Regulation (EU) No 596/2014 and Commission Delegated Regulation (EU) 2016/958, and to supervision by the Comissão do Mercado de Valores Mobiliários (CMVM) in Portugal. Supervision does not constitute regulatory approval or endorsement of the research.

The article reviews research published on 7, 21 and 28 August 2026, together with excerpts subsequently released on VMF’s Market View. Original analyses and portfolio decisions retain their respective research dates; subsequent developments are identified separately. This is a selective guide, not a complete investment thesis, a comprehensive performance record or confirmation that earlier recommendations remain unchanged. The underlying publications provide the relevant sources, methodologies, assumptions, valuation work, risks and dated disclosures.

The content is not personalised investment advice and does not consider any reader’s objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. It does not constitute portfolio management, order execution or an offer to transact in any financial instrument. References to other investors, institutions and their decisions are analytical case studies, not endorsements or evidence of affiliation. This article also promotes VMF Research’s paid publications.

The Dovish Shock, TACO Zone and Abundance Shock are conditional analytical frameworks, not assurances of policy decisions, technological progress or investment outcomes. Historical relationships may change, and the TACO Zone is not a permanent yield threshold or promise of intervention. Views are reviewed through monthly publications, with weekly or ad hoc updates where appropriate.

Model Portfolios are illustrative research portfolios, not client accounts, investable funds or transactions executed by VMF Research. Unless otherwise stated, model returns use published reference prices, include recorded cash distributions and exclude taxes, transaction costs, custody charges, bid and ask spreads and currency effects. Past performance, model results and scenarios are not reliable indicators of future returns. Investments can lose some or all their value. Precious metals, mining equities and crypto exposures can be particularly volatile. Diversification does not guarantee capital protection.

Disclosure of Interests: the original August disclosures reported relevant long positions held by legal entities controlled by the author in the KraneShares CSI China Internet ETF, ARK Innovation ETF, Altius Minerals Corporation, Universal Music Group N.V. and Amrize Ltd., together with listed ETPs providing exposure to Bitcoin, Ether and Solana. These financial interests may create conflicts and may have changed since those disclosures. Their existence does not validate the analysis, reduce risk or imply suitability.

Neither VMF Research nor the author received compensation from covered entities for preparing the original research, and no covered entity reviewed, approved or amended its conclusions before first dissemination. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser.

Investors waiting for the Fed to blink may be watching the wrong building.

In The Hawk and the TACO Zone, we examined how rising borrowing costs could constrain Washington’s ambitions. The Treasury’s August decision to expand bond buybacks gave that argument a new chapter, and our Dovish Shock thesis another policy channel to watch.

Stanley Druckenmiller’s objection makes the story more interesting still. A government responding to pressure from its creditors may discover that the response itself changes what those creditors demand.

This excerpt was first published for paid Alpha Tier subscribers on 28 August 2026. It examines the policy developments behind that tension and their significance for the financial conditions we have been tracking for months.

The cross-asset analysis and Model Portfolio implications remain exclusive to subscribers, including the harder decision: whether improving evidence justified adding exposure or whether the market had already done enough of the work.

Good reading.

The long bond has been warning Washington for months.

We first introduced the TACO Zone earlier this year as shorthand for the point at which rising long-term Treasury yields cease to be merely a market price and begin to interfere with the political economy around them. Mortgage rates rise. Corporate financing becomes more expensive. Equity duration compresses. Interest expense accelerates. And an administration attempting to finance tax cuts, defence, industrial policy and an AI infrastructure boom discovers that the bond market has its own vote.

The exact level was never sacred.

But the behaviour around it was...

That is why we returned to the framework in August’s VMF’s Strategic Asset Allocation. The 30-year Treasury had again been pressing against the 5.10–5.20% region, and our conclusion was that the long end was becoming an increasingly important constraint on both fiscal and monetary freedom.

Then it pushed through.

On 19 August, the 30-year yield traded as high as roughly 5.34% intraday, close to levels last seen almost two decades ago. Treasury responded the same day. Beginning in September, liquidity-support buybacks in the 10-to-20-year and 20-to- 30-year sectors will rise from a maximum of $2 billion to at least $4 billion per operation. Treasury explicitly framed the change as additional support for liquidity in the long end.

We should take the Treasury at its word about the mechanics. This is not QE. It is not formal yield-curve control. Treasury has not announced a target for the 30- year yield, and Bessent has since confirmed that regular auctions, including long- dated issuance, will continue.

But the announcement answered the question we actually cared about.
The long bond produced a policy response.

That is a more important development than whether $4 billion is large relative to a Treasury market measured in tens of trillions.

It is not.

But the signal is...

The market initially treated the announcement exactly as one would expect if the government had become less tolerant of rising long-term financing costs: long yields fell, the dollar weakened and both gold and Bitcoin moved higher.

For our Dovish Shock, that is significant.

We introduced the thesis because the market had reduced Kevin Warsh to a caricature: structurally hawkish, almost irrespective of the incoming data. Our argument was narrower and, we think, more useful. Warsh does not need to become a dove. He merely needs to deliver a policy path that is less restrictive than the path already embedded in asset prices.

That thesis has gained strength.
Not marginally.
Materially.

You might also like reading:

The reason is that the easing impulse we envisaged no longer depends entirely on Warsh. There is now a second actor... the Treasury.

And the Treasury controls something we have spent a considerable amount of time analysing in these pages: the Treasury General Account (TGA).

This brings us back to our 2025 work on Fed Net Liquidity.

Our shorthand has always been deliberately simple: Federal Reserve assets, less the TGA, less the overnight reverse-repo balance.

It is not an official Federal Reserve statistic. It is a way of tracking the major balance-sheet channels through which dollars are either available to the private financial system or absorbed from it.

For several years, the reverse-repo facility acted as a huge liquidity reservoir. At its December 2022 peak, more than $2.5 trillion sat at the ON RRP facility. As Treasury issued bills and money-market funds migrated into them, that stock of cash was released back towards private markets and cushioned the liquidity effect of quantitative tightening.

That reservoir is now effectively empty.

On 21 August, domestic ON RRP usage was just $0.2 billion.

This changes the map.

There is no longer a multi-trillion-dollar RRP buffer capable of absorbing Treasury issuance while simultaneously releasing liquidity into the system. In this context, the TGA has become the dominant discretionary swing factor in our Net Liquidity framework.

And it is large.

The Treasury’s account at the Fed stood at approximately $936 billion on 19 August, after reaching roughly $970 billion in late July.

A dollar accumulated in the TGA is, all else equal, a dollar withdrawn from bank reserves. A dollar spent out of the TGA moves in the opposite direction.

This is not our interpretation of obscure financial plumbing. The New York Fed described precisely the same mechanism earlier this year when reviewing 2025: the debt-ceiling-driven TGA drawdown temporarily increased “system liquidity”, defined there as reserves plus ON RRP (the subsequent rebuilding of the account reversed that support).

We lived through that liquidity cycle in our Model Portfolios last year.

Now the lever is back in focus.

Bessent said this week that the enlarged TGA could be used to fund Treasury buybacks rather than requiring equivalent new short-term issuance. No amount has been committed, and we should not convert an available option into a forecast. But the fact that the Treasury Secretary is publicly discussing the TGA as a potential funding source changes the liquidity conversation.

A TGA drawdown would not be QE.

It would, however, release cash currently sitting idle at the Federal Reserve back into the financial system.

For markets, labels are less important than flows.

This is where our view has become more assertive.

When we first presented the Dovish Shock, we were describing an asymmetry. We are now beginning to observe the transmission mechanism.

Warsh remains the central monetary-policy variable. But the Administration is showing that it is unwilling to treat the long end as a passive spectator. The Treasury has increased buybacks precisely where pressure has been greatest.

Bessent has opened the possibility of using the TGA. The dollar has responded. Scarcity assets have responded. And the RRP buffer that obscured parts of the liquidity signal in previous years is no longer there.

The probability of our Dovish Shock has increased.

That does not mean its path will be clean.

Stanley Druckenmiller has just articulated the strongest argument against what Treasury is doing, and it deserves considerably more attention than a footnote. Druckenmiller called the expansion of long-end buybacks a “mistake” that risks costing the United States credibility in its own bond market. His objection is not primarily about $4 billion.

It is about precedent.

Once the Treasury begins responding to an inconvenient bond price, how does the market distinguish liquidity management from price management?

And if 5.3% is uncomfortable enough to produce larger purchases, what happens at 5.5%?

Or 6%?

Does $4 billion become $8 billion?
Does the maturity range widen?
Does the TGA become a recurring source of intervention?

Druckenmiller’s warning is that the Treasury market’s greatest asset is precisely the thing that cannot be manufactured through buybacks: confidence that price discovery remains credible. He argues that the durable way to lower long-term rates is to address the primary deficit, not to lean against the market that is pricing it.

That criticism carries unusual weight.

Druckenmiller is not an academic observing the bond market from a distance. He built one of the great macro investing records by treating currencies, rates and liquidity as parts of the same system. He also knows both Bessent and Warsh personally and professionally. When someone with that background warns that a tactical liquidity measure risks becoming a credibility problem, we should listen.

But his objection does not weaken our Dovish Shock thesis.
It sharpens it.

There are now two possible paths.

The benign version is the one markets initially embraced: Treasury improves long-end functioning, Warsh proves less restrictive than expected, the TGA releases liquidity, real yields soften and the dollar loses support.

The more dangerous version is that investors conclude Washington has become too sensitive to bond prices, forcing a larger fiscal premium into the long end.

Those paths are very different for duration.

They are not necessarily as different for monetary scarcity.

Gold can benefit from easier liquidity. It can also benefit from declining confidence in sovereign financial discipline. Bitcoin and Ethereum are less directly exposed to real yields than gold, but more sensitive to the liquidity, dollar and risk-appetite consequences that often accompany changes in them. That is one reason we refuse to treat all of our Alternatives positions as the same trade.

That distinction will become central in the next section.

There is one final reason the current setup deserves our attention: the Dovish Shock never required an emergency easing cycle. That remains one of the most misunderstood parts of the thesis. A market positioned for hawkish policy can reprice violently when it receives less hawkishness, even if the absolute stance of policy remains restrictive.

The same logic now extends beyond the Fed.
Treasury does not need to cap the 30-year yield.
The TGA does not need to fall by $500 billion.
Warsh does not need to slash rates.

A series of smaller changes can point in the same direction:

• less pressure from the policy rate;

• more support at the long end;

• a declining TGA;

• lower real yields;

• a softer dollar;

• and a turn in liquidity.

The ingredients are no longer hypothetical. Some are already beginning to appear. That is why we are more confident in the Dovish Shock today than we were when we first published it.

And markets that are particularly sensitive to those ingredients have begun behaving accordingly. The question now is whether those moves are merely confirming what we already own… or whether the evidence has become strong enough to justify a change in positioning.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces “The Treasury Just Blinked” from the August 2026 issue of Alpha Tier. The underlying research, charts and Model Portfolio references are stated as of 28 August 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Alpha Tier Model Portfolio is an illustrative research portfolio and does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account. The analysis combines macroeconomic, monetary-policy, fixed-income, liquidity and balance-sheet research within a medium- to long-term framework. The Dovish Shock and TACO Zone are conditional analytical frameworks, not assurances of particular policy or market outcomes. The TACO Zone does not represent a permanent yield threshold. Net Liquidity is an analytical proxy used by VMF Research, not an official Federal Reserve measure. Views are reviewed through monthly publications, with weekly or ad hoc updates where appropriate.

Treasuries, precious metals, cryptoassets and related financial instruments carry differing market, interest-rate, currency, liquidity, custody and counterparty risks. Cryptoassets also involve substantial technological, cybersecurity and regulatory risks. Exchange-traded products introduce product-specific dependencies that differ from direct ownership of the underlying asset. Investors may incur substantial losses, and policy intervention does not provide capital protection.

Disclosure of Interests: as of the research cut-off, legal entities controlled by Vasco Marques de Freitas held long positions in listed exchange-traded products providing economic exposure to Bitcoin and Ether, both discussed in this article. These holdings may benefit from increases in the value of the underlying assets and constitute financial interests and potential conflicts of interest. Readers should consider them when evaluating the analysis. Their existence does not validate its conclusions, reduce investment risk or imply suitability for any reader.

Neither VMF Research nor the author received compensation from any issuer covered in connection with the preparation of the original research, and no issuer reviewed, approved or amended its investment conclusions before first dissemination. Past performance, Model Portfolio performance, historical market relationships and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

Tether has a peculiar way of preparing for the future of money: it keeps buying gold.

The company behind USDT, the dollar-linked stablecoin, earns interest on a vast portfolio of US government securities. It also allocates capital to Bitcoin and physical bullion. For a business making substantial profits from interest-bearing reserves, choosing an asset that pays nothing deserves a closer look.

In The Only Assets That Owe You Nothing, we argued that gold and Bitcoin could perform complementary jobs within a portfolio. Tether offers a useful case through which to revisit that argument, especially for readers accustomed to hearing that the success of one must come at the expense of the other.

Much of our recent research has examined what happens when a useful investment rule becomes a substitute for judgement. The debate over money provides fertile ground. Investors can spend years defending a monetary allegiance without asking whether it serves the portfolio they actually need.

Tether’s choices invite that question. They also raise a less comfortable one: how much can we learn about an investment from the asset alone, without examining the structure surrounding it?

This excerpt was first published for paid Alpha Tier subscribers on 28 August 2026. The full issue follows the argument into our Dovish Shock thesis, Treasury policy and the market signals behind the decision to maintain the Model Portfolio’s positioning rather than chase the recovery.

Good reading.

The most interesting thing about Tether’s gold purchases is not that a crypto company likes gold.

It is that Tether apparently sees no reason to choose.

For years, one of finance’s strangest tribal disputes has pitted two assets with more in common than either camp often admits against each other. Gold advocates dismiss Bitcoin as speculative code with no history. Bitcoin maximalists describe gold as monetary technology whose best days disappeared with the telegraph.

The argument makes for excellent social media.

It makes considerably less sense as portfolio construction.

Tether offers an unusually useful test case because this is not an institution arriving reluctantly from traditional finance. It was built inside the crypto economy. It operates USDT, the dominant dollar-linked stablecoin, invests heavily in US government securities and has direct exposure to Bitcoin and blockchain infrastructure. If anyone should be predisposed to believe that digital scarcity has made physical monetary scarcity obsolete, Tether would be a reasonable candidate.

Instead, it has become one of the most aggressive private buyers of gold in the world.

Tether International purchased approximately 27 tonnes during the first half of 2026, taking its physical gold holdings to roughly 150 tonnes by June. The company has separately built Tether Gold, or XAU₮, whose tokens represent direct ownership of physical bullion held in Swiss vaults. At the end of Q2, that product was backed by another 22 tonnes of gold, although those bars belong economically to XAU₮ holders rather than to Tether itself.

That distinction is important.

So is the scale.

Tether is not buying a few bars to decorate a balance sheet.

At the beginning of the year, CEO Paolo Ardoino said the company considered it reasonable to hold roughly 10% of its investment portfolio in Bitcoin and 10–15% in physical gold. Tether had already been buying gold at a pace of around two tonnes per week.

Read those numbers again.

Bitcoin and gold.
Not Bitcoin instead of gold.
Not gold because Bitcoin failed.
Both.

That revealed preference is more useful than another thousand-page debate about which asset constitutes “real money”. A sophisticated allocator does not need assets to belong to the same philosophical camp. It needs them to perform useful economic functions.

Gold and Bitcoin overlap in one very important respect. Neither is somebody else’s liability. That was the argument behind June’s The Only Assets That Owe You Nothing. But overlap is not equivalence. Gold has no technological obsolescence risk. Its monetary history is measured in millennia. Central banks already accept it as a reserve asset, its physical supply responds slowly to price and no network needs to remain operational for a bar in a vault to continue existing. Bitcoin offers something gold cannot: digitally verifiable scarcity that can move globally without moving a physical object. Its supply rules are transparent, settlement is native to the network and ownership can cross borders at a speed and cost that bullion cannot reproduce.

You might also like reading:

The strengths of one do not erase the strengths of the other.

They are different answers to a partially shared problem.

And that is precisely why owning both can be more rational than choosing a side.

There is another reason Tether’s behaviour deserves attention.

Its core business gives it access to something gold does not provide: yield.

USDT reserves are heavily invested in short-duration US government securities. As of the first quarter, Tether reported approximately $141 billion of direct and indirect Treasury exposure. Those securities pay interest. Gold does not.

That makes the decision to accumulate bullion more revealing, not less. A Treasury bill offers a contractual yield backed by the US government. A gold bar offers no coupon whatsoever.

For an institution earning billions from the spread generated by a vast pool of dollar reserves, allocating meaningful capital to an asset with no cash flow amounts to accepting an obvious opportunity cost.

Why do it?

Because carry is only one property of money.

Liquidity matters.
Credit quality matters.
Duration matters.

And in a world of rising sovereign indebtedness, geopolitical fragmentation and increasingly active monetary and fiscal authorities, independence from the issuer can matter too.

This does not make Treasury bills dangerous. Tether continues to own them in enormous size, and for good reason.

It makes the composition interesting.

The same institution can rationally want highly liquid US government paper for the operating demands of a stablecoin, Bitcoin for digital scarcity and gold for physical monetary scarcity.

Three assets. Three different jobs. One balance sheet.

That is a much more sophisticated way of thinking about portfolios than asking which one is destined to “win”.

And Tether is going further than simply accumulating bullion.

In February, it invested $150 million for roughly 12% of Gold.com, with plans to integrate XAU₮ into the platform and explore the purchase of physical bullion using digital dollars. XAU₮ has meanwhile expanded across blockchain infrastructure and received regulatory recognition as an accepted spot commodity within Abu Dhabi Global Market.

This is where the story becomes more interesting. Tether is not merely putting old money on a new balance sheet. It is attempting to put old money on new rails.

There is an important conceptual point here.

Technology usually does not eliminate scarce assets simply because it improves how claims on those assets can move. The internet did not eliminate warehouses because commerce became digital. Electronic trading did not eliminate securities because exchanges became faster. And blockchain need not eliminate gold merely because value can now settle digitally.

In some cases, technology can make the underlying scarce asset more monetisable.

Tether Gold is an early example. A London Good Delivery bar remains physical, costly to produce and impossible to replicate with software. Blockchain changes the ownership and settlement layer around it. One full XAU₮ represents one fine troy ounce of allocated physical gold, while the token can be transferred on-chain without physically moving the bar each time ownership changes.

That is not gold losing to crypto. It is crypto making gold more crypto-like. And it points towards a broader principle we have repeatedly encountered in our research.

Abundance does not necessarily destroy scarcity. It can increase the economic value of the scarce thing by making it easier to access.

This is why the Gold-versus-Bitcoin debate increasingly strikes us as the monetary equivalent of arguing whether investors should own software or electricity.

The assets perform different functions.
Their risks are different.
Their return distributions are different.

There will be environments in which one materially outperforms the other. That is exactly why a portfolio may want both.

Strategy’s ( recent Bitcoin sales provide a useful counterpoint. The company has not suddenly repudiated Bitcoin. It has built an increasingly complex corporate structure around it, complete with debt, preferred securities, distributions and liquidity requirements. When Strategy sold 1,638 Bitcoin recently, part of the proceeds went towards preferred-stock obligations and repurchases.

The lesson is not that Michael Saylor changed his mind. It is simpler...

Bitcoin owes nobody anything. Strategy does.

Place a scarce asset inside a financed wrapper and the behaviour of the wrapper can eventually dominate the behaviour of the asset. We encountered the same principle in Tier Two with Terry Smith. Fundsmith may hold businesses capable of recovering over the long term, but an open-ended fund faces redemption risk that Berkshire Hathaway does not. The liability structure determines how much time the asset owner actually possesses.

Tether is interesting from the opposite direction. Its balance sheet contains enormous liabilities in the form of redeemable dollar tokens, so liquidity remains essential. But once those obligations are supported by ample short-duration reserves, the company has increasingly directed its economic surplus towards assets whose supply cannot be expanded by a central bank.

Gold.
Bitcoin.
Not one against the other.
Both… against a different problem.

That is the part of Tether’s behaviour we think investors should pay attention to.

The old debate asks which form of scarcity will replace the other. The better question is why some of the most sophisticated capital inside digital finance increasingly sees value in owning more than one form of scarcity at the same time.

Because the market is now beginning to do the same thing. Gold, silver, Bitcoin and Ethereum have recently started moving with an unusual degree of sympathy. And the explanation may have less to do with a sudden outbreak of monetary philosophy than with something happening several steps upstream.

In the market that prices the world’s most important financial liability.

US Treasuries.

That is where we go next.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces “The End of Simple Frameworks” from the August 2026 issue of Alpha Tier. The underlying research, charts and Model Portfolio references are stated as of 28 August 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 28 August 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Alpha Tier Model Portfolio is an illustrative research portfolio and does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines monetary, fundamental, balance-sheet and portfolio-construction research within a medium- to long-term framework. Conclusions concerning monetary scarcity, the complementary roles of gold and Bitcoin, tokenisation and the effects of financing structures are analytical judgments rather than assurances. Corporate allocation decisions are examined as case studies, not as allocations suitable for every investor. Views may change and are reviewed through monthly publications, with weekly or ad hoc updates where appropriate. Precious metals, cryptoassets and related instruments carry differing market, liquidity, custody, counterparty, regulatory and technological risks. Stablecoins may lose their intended reference value. Tokenised bullion and exchange-traded products introduce legal, custody, operational and redemption dependencies that differ from direct ownership of the underlying asset. Corporate crypto exposure can introduce additional leverage, refinancing, distribution and dilution risks. Combining assets does not guarantee diversification or capital protection, and investors may lose some or all of their capital.

Disclosure of interests: as of the research cut-off, legal entities controlled by Vasco Marques de Freitas held long positions in listed exchange-traded products providing economic exposure to Bitcoin and Ether. These holdings may benefit from increases in the value of the assets discussed and constitute financial interests and potential conflicts of interest. Readers should consider them when evaluating the analysis. Their existence does not validate its conclusions, reduce investment risk or imply suitability for any reader.

Neither VMF Research nor the author received compensation from any issuer covered in connection with the preparation of the original research, and no issuer reviewed, approved or amended its investment conclusions before first dissemination.

Terry Smith spent fifteen years telling investors that most of their activity was a tax on compounding.

Then, in 2026, he turned over more than half his portfolio in six months.

For a manager whose philosophy had been distilled into three instructions (buy good companies, don’t overpay and do nothing) that was not routine portfolio maintenance. Something in the calculus had changed.

When a great investor changes a rule that helped make him successful, the interesting part is rarely the trade itself. It is the world that forced the rethink.

That became our starting point for August’s VMF’s Security Selection.

Quality investing has always relied on patience. A strong franchise hits trouble, the market extrapolates the weakness and the patient investor gives management time to repair what others have already written off.

The trouble is that time may no longer be as generous as it once was.

AI is compressing development cycles, lowering the cost of experimentation and giving smaller competitors access to capabilities that previously required large teams, deep pockets and years of accumulated know-how. A company can still have a recognised brand, recurring revenue and attractive historical returns on capital while the economic ground beneath those numbers is moving faster than management can respond.

That is why we called the issue:

Quality Has a Half-Life.

The phrase is deliberately provocative. We are not suggesting that great businesses suddenly expire on schedule. We are asking whether the durability investors once inferred from yesterday’s economics now needs to be re-underwritten more frequently.

The question we kept coming back to was simple:

Would this company become more valuable or less valuable if intelligence became ten times cheaper?

Following that question took us into an industry where intelligence is extraordinarily valuable, but where intelligence alone is nowhere near enough.

Life sciences.

A new model can analyse data, draft documents and accelerate research. It cannot, by itself, make a clinical process compliant, maintain an audit trail, satisfy a regulator or turn a scientific insight into an approved medicine.

Between raw intelligence and a usable outcome sits an enormous layer of specialised workflows, permissions, data, trust and institutional knowledge.

That is where we went looking.

The business we ultimately recommended to Tier Two subscribers does not manufacture semiconductors, train foundation models or depend on discovering the next blockbuster drug. Its position is further downstream, inside the operating machinery of the pharmaceutical and biotechnology industry.

And the more capable AI becomes, the more interesting that position may become.

August’s paid issue contained the company, our full underwriting, valuation work, scenario analysis, risks and the initial Model Portfolio allocation.

We are keeping the recommendation behind the paywall this month.

What follows instead is the thinking that led us to it: why Terry Smith’s extraordinary change in behaviour caught our attention, why the old Quality playbook may need another test, and why time has quietly become one of the most important variables in security selection.

This section was first published to paid subscribers in VMF’s Security Selection on 21 August 2026.

Smith changed the rule.

We wanted to understand the reason.

The answer eventually led us to a business built for the place where cheap intelligence meets expensive consequences.

Good reading.

Changing a process that has failed is easy to explain.

Changing one that made you successful is something else entirely.

Terry Smith has spent much of his career reading what other investors ignored. He began at Barclays in the 1970s, became one of London’s leading bank analysts during the 1980s and later headed UK company research at UBS Phillips & Drew. In 1992, he published Accounting for Growth, a forensic examination of the accounting techniques companies used to make weak economics appear stronger.

The book became a bestseller. It also contributed to his dismissal.

That episode established the public character Smith would carry through the rest of his career: direct, sceptical, willing to challenge management and unusually alert to the difference between reported growth and genuine economic value creation.

More importantly, it established his willingness to defend an analytical conclusion even when the immediate professional cost was substantial.

Smith subsequently joined Collins Stewart, led its management buyout and

flotation, and helped build Tullett Prebon into one of the world’s largest interdealer brokers.

Then, in 2010, at an age when many successful executives begin preparing for retirement, he launched Fundsmith.

The investment proposition was deliberately narrow. Fundsmith would own a concentrated collection of companies capable of sustaining high returns on operating capital. Their competitive advantages would be difficult to replicate. They would not require substantial leverage to produce attractive returns. They would possess long and relatively predictable opportunities to reinvest their cash flows. They would also need to be resilient to change, particularly technological change.

The fund would then avoid the activity through which managers routinely dilute those advantages. There would be no market timing, no index hugging and no trading for the sake of appearing busy.

The philosophy was compressed into three instructions: buy good companies, do not overpay and do nothing.

The logic was that compounding requires inactivity.

A company that earns 25% on capital and can reinvest a meaningful proportion of its cash at the same rate does not need a constantly changing macroeconomic forecast to create value. Its internal economics perform the compounding. The investor’s principal task is to avoid interrupting it.

Low turnover reduces commissions, spreads and taxes. It avoids replacing a business the manager already understands with one that must be learned from the beginning. It also reduces the probability that emotion will transform temporary price volatility into a permanent loss of capital.

Most importantly, inactivity allows the holding period itself to become an advantage.

The market is generally efficient at discounting next year’s earnings. It is less comfortable underwriting what a high-return business may become after ten years of reinvestment. The further the investment horizon extends, the more difficult it becomes for conventional forecasts, quarterly benchmarks and career incentives to compete.

“Do nothing” was therefore not the absence of a process.

It was the final expression of the process.

That is what makes Smith’s latest change so important. But to appreciate its significance, Fundsmith’s track record must be divided into two very different periods.

During the first, Smith did not merely outperform. He established one of the most impressive records in European fund management. From the fund’s launch in November 2010 through the end of 2021, Fundsmith returned 570.7%, compared with 287.1% for the MSCI World Index in sterling. An initial investment of £100 became approximately £671 in Fundsmith and £387 in the index. On an annualised basis, the fund compounded at 18.6%, almost six percentage points faster than global equities.

That was not marginal outperformance. It was the record on which Smith’s reputation (and the commercial success of Fundsmith) was built.

Then the pattern broke...

From the beginning of 2022 through June 2026, Fundsmith produced a return of only approximately 3.3%. Over the same period, the MSCI World Index in sterling advanced by roughly 63%.

The first period turned £100 into £671 while the index reached £387.

During the second, £100 invested in Fundsmith became approximately £103 while the same amount invested in the index became £163.

By June 2026, Fundsmith could still report an excellent since-inception record. The fund had returned 592.6% cumulatively and 13.1% annually, compared with 530.9% and 12.5% for the MSCI World. But those figures increasingly concealed the change beneath them. Almost all the relative wealth creation had occurred during the first part of the record. The subsequent four and a half years had consumed most of the lead.

The track record was no longer one continuous story.
It had become two.

This distinction matters because Smith did not alter his process after one disappointing quarter. Fundsmith had entered a sustained period of relative underperformance.

The concentration of index returns in a small group of large technology companies created an obvious headwind. A weaker US dollar reduced the sterling value of a portfolio with substantial American exposure. Several traditional Quality holdings also experienced company-specific disappointments that proved deeper or more persistent than expected.

The fund had marginally underperformed in 2021 and then fell progressively further behind through 2022, 2023, 2024 and 2025...

...then, the first half of 2026 brought the issue to a head.

Fundsmith declined by 2.9%, while the MSCI World Index in sterling gained 11.2%. A gap that had already persisted for several years widened by another 14.1 percentage points in six months.

Smith responded with activity on a scale the fund had never previously displayed.

Portfolio turnover reached 51.8% during the half-year.

The fund opened twelve positions and exited, or began exiting, thirteen. Companies including Nike, Novo Nordisk, LVMH, Unilever, Zoetis, Coloplast and Wolters Kluwer were sold or placed in the process of being sold. New capital moved into businesses including AppLovin, GE Vernova, Legrand, Mastercard, Netflix, Nextpower, Sage, TJX, TSMC, Uber and Yum! Brands.

The composition of the portfolio changed.

So did its operating instructions.

Smith said Fundsmith would remain committed to owning good businesses at reasonable valuations. The first two rules survived. The third required modification...

The fund would become more active.

It would take greater account of both fundamental and share-price momentum. Most strikingly, it would become considerably less willing to use one of the established techniques of Quality investing: purchasing a great company after what appeared to be a temporary setback.

Smith compared that technique to catching a falling knife.

Fundsmith’s fingers had been cut too often.

Smith’s explanation focused heavily on the structure of the modern market. Passive vehicles now represent a large share of equity ownership, but their influence over marginal prices may be considerably larger than their share of assets suggests. Index flows purchase securities according to their weight rather than their valuation. Rising prices increase those weights. Strong performance attracts further flows. The process can become self-reinforcing, directing more capital towards the securities that have already appreciated while depriving neglected parts of the market of marginal demand.

That does not make index investing irrational...

For most investors, low cost, diversification and behavioural discipline remain formidable advantages. But it does mean that the resulting market is not passive in the ordinary meaning of the word. Somebody determines the index, and capital subsequently responds to that decision and to the price movements that follow.

An active manager can believe that this process has become detached from fundamentals and still lose assets while waiting for the detachment to reverse.

The overlooked issue therefore sits on the liability side of the portfolio.

An investment portfolio does not exist independently of the capital structure attached to it. Warren Buffett could purchase American Express during the salad-oil scandal because the structure surrounding his capital gave him the ability to wait. Berkshire Hathaway was not required to redeem nervous investors at the bottom.

An open-ended fund is different.

Its investors can withdraw capital precisely when the manager believes the opportunity is becoming most attractive.

That creates a dangerous mismatch. The investment thesis may require patience, while the funding base refuses to provide it.

Smith’s decision to incorporate momentum can therefore be interpreted as risk management at the fund level rather than the abandonment of fundamental investing. He is attempting to avoid a situation in which the portfolio is eventually proved right only after redemptions have made that vindication irrelevant.

This distinction matters for subscribers to VMF Research.

Our Model Portfolios do not face daily redemptions. They are not forced sellers. An investor with permanent capital can tolerate a period of unpopularity that an open-ended manager may not survive commercially. We should not surrender that advantage by importing a process designed to solve somebody else’s liability problem.

But neither should we dismiss Smith’s message.

It leaves us with an uncomfortable question: is this adaptation, or surrender?

If a manager purchases only after fundamental and price momentum improve, he risks selling near the bottom and returning after the recovery has become visible. If fund flows begin influencing security selection, the liabilities start dictating the assets rather than the other way around. And if a Quality investor responds to several years of underperformance by buying more of what the market already rewards, adaptation can quietly become style drift.

Some of Fundsmith’s new holdings make the question legitimate. TSMC, AppLovin and GE Vernova participate directly or indirectly in themes that Smith has also described in the recent past as dangerously momentum-driven. Purchasing them after substantial appreciation may ultimately prove prescient. It may also suggest that avoiding a popular trade became more difficult as the commercial cost of not owning it increased.

The evidence is not yet sufficient to decide...

Smith may be adapting rationally to a new market structure. He may be weakening a valuable discipline at precisely the wrong stage of the cycle. Those two interpretations can remain simultaneously plausible until performance and future portfolio behaviour separate them.

Our purpose is not to defend or prosecute him.
It is to extract the investment lesson.

Fundsmith’s recent performance cannot be explained exclusively by passive flows, index concentration or an unfavourable currency. Low-cost Quality indices have also outperformed the fund over several recent periods. Security selection, sector exposure, valuation and the specific companies Fundsmith chose to own all contributed. Any explanation that attributes every disappointing result to the structure of the market would be too convenient.

Smith appears to recognise this. Fundsmith sold companies for reasons including weak organic growth, management failure, excessive valuation and deteriorating competitive conditions. The revised process is not simply an instruction to buy whatever is rising. Fundamental momentum remains central.

The more revealing change is Smith’s reduced willingness to assume that a setback is temporary.

That is where his experience connects with a much broader problem confronting every Quality investor.

The classic Quality opportunity emerges when an excellent company encounters a recoverable problem. The market extrapolates the weakness. The patient investor looks through it. The business repairs the damage, its high returns reappear and the valuation normalises.

Warren Buffett’s investment in American Express during the salad-oil scandal remains the canonical example. Fundsmith’s purchase of Microsoft near the end of Steve Ballmer’s tenure offers a more recent one.

The method works only if three conditions hold.

The franchise must remain intact. Management must possess the capacity and determination to repair the problem. And the company must have enough time to do so.

Technology is placing unprecedented pressure on the third condition.

Artificial intelligence is reducing the cost of producing software, analysing information, designing products, generating content, serving customers and testing new commercial ideas. Competitors can move more quickly. Customers can evaluate alternatives more easily. Small organisations can assemble capabilities that previously required years of investment, large teams and substantial amounts of capital.

This is not merely a software problem.

Cheaper intelligence can alter research, manufacturing, distribution, marketing, customer service, financial analysis, product development and corporate decision-making across almost every industry. It can strengthen an incumbent by increasing its productivity. It can also give a competitor the tools required to attack an advantage that once appeared prohibitively expensive to replicate.

A management team that requires three years to complete a turnaround may discover that the market it planned to re-enter no longer exists in the same form. The weakness may still be repairable in theory while becoming irrelevant in practice.

This does not mean that every earnings disappointment has become permanent or that patient investing no longer works.

It means the cost of misclassification is rising.

The financial statements will often reveal the change late. Revenue can remain stable while customer engagement deteriorates. Margins can improve temporarily because a company is underinvesting in its product, distribution or people. Buybacks can support earnings per share while the strategic position weakens. A familiar brand can preserve yesterday’s cash flows even as tomorrow’s relevance migrates elsewhere.

A damaged share price can recover.
A damaged competitive position may not.

Quality investors must therefore look beyond historical returns on capital. They must determine whether the forces responsible for those returns are still strengthening... or whether the reported numbers are preserving the appearance of a franchise whose economic relevance has already begun to fade.

This framework changes what we should look for.

• What is the customer really paying for and is that source of value becoming scarcer or easier to replicate?

• Which part of the industry’s economics will control bargaining power as

  technology changes?

• Does the company own a brand, distribution network, cost advantage,

  licence, patent portfolio, physical asset, proprietary dataset, installed base or customer relationship that becomes more valuable as capabilities spread?

• Can it capture the productivity gains created by better technology, or will competition force it to pass those gains entirely to customers?

• Does its balance sheet provide enough room to invest before the legacy business begins weakening?

• Is management reallocating capital towards the next source of profit, or

  protecting the economics of the previous one?

• Would the company become more or less relevant if intelligence became

  ten times cheaper?

These questions apply to a consumer brand as much as to a software platform; to an industrial manufacturer as much as to a pharmaceutical company; and to a financial exchange as much as to a retailer. The relevant source of control will differ. It may reside in intellectual property, physical infrastructure, distribution, regulation, scale, trust, data, customer access or an ecosystem that becomes more valuable with every additional participant.

What matters is whether the company controls something the next competitive regime will make more valuable... not merely something the previous regimerewarded.

These questions do not replace the traditional Quality Framework.
They make it harder to satisfy.

High returns on capital, recurring revenue, pricing power, strong balance sheets and long reinvestment runways remain essential. But historical excellence is no longer sufficient evidence of future durability. The investor must also understand how the source of that excellence behaves when intelligence becomes cheaper, competitive responses become faster and the time available to correct a strategic error contracts.

The best businesses in the next phase of the AI cycle may not be those most visibly associated with artificial intelligence. They may be companies that control the scarce assets, trusted relationships or essential processes through which cheaper intelligence must pass before it can create economic value.

That is why one of Fundsmith’s newest positions matters to us.
Not because Smith owns it.
And not because his process change provides our catalyst.

It matters because the company may offer an answer to the question his pivot has raised.

This company is not merely another Quality company caught in the software selloff. It represents what the next generation of Quality looks like…

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the August 2026 issue of VMF’s Security Selection. The underlying research was completed on 21 August 2026 at 4:00 p.m. Eastern Daylight Time and first disseminated to paid subscribers on 21 August 2026 at 9:00 p.m. Eastern Daylight Time.

The article discusses investment philosophy, Quality investing, portfolio construction, market structure, technological disruption and the distinction between temporary operational setbacks and permanent competitive impairment. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk.

References to Terry Smith, Fundsmith and individual securities bought, sold or discussed in connection with Fundsmith’s portfolio are included for analytical and historical purposes. They describe Fundsmith’s investment process and its evolution and should not be interpreted as recommendations by VMF Research to buy, sell or hold those securities, nor as implying any affiliation with or endorsement by Terry Smith or Fundsmith.

The discussion of artificial intelligence, competitive advantage, Quality investing and the potential shortening of the time available for companies to respond to technological disruption reflects VMF Research’s analytical framework and judgment. Competitive positions may evolve differently from those contemplated, technological adoption may proceed faster or slower than expected, and businesses considered durable today may experience unforeseen changes in regulation, customer behaviour, technology or industry structure.

The article refers to a new Tier Two recommendation contained in the original August publication but does not identify or reproduce that recommendation, its valuation, Model Portfolio allocation or associated investment conclusions. Paid subscribers received the complete analysis, including the relevant assumptions, valuation scenarios, principal risks and portfolio implementation.

Neither VMF Research nor the author received compensation from any issuer, investment manager or other entity discussed in connection with the preparation of this research, and no covered entity reviewed, approved or amended its conclusions before first dissemination.

Past performance, historical investment records, Model Portfolio performance and examples of successful or unsuccessful investment strategies are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

Biotechnology was moving forward. New pockets of strength were beginning to emerge. And the composition of our Top 10 was changing with them.

Markets have rotated again.

With prices through September 2, the Debasement Trade has come roaring back.

Here’s our updated Top 10 Positions - September 2nd 2026:

Junior gold miners are now showing a 198% total return from our original reference price. Altius Minerals is sitting at roughly the same level. Senior gold miners are up 177%. Silver is back above 100%. And our broader metals and mining exposure has returned to the Top 10.

Quite a reversal from July.

But the more interesting development is not simply that gold-related assets are working again.

It is why they may be working again.

Earlier this year, we introduced what we called the TACO Zone: the region in long-term Treasury yields where rising borrowing costs would stop being merely a market price and begin interfering with Washington’s own agenda.

We later showed how movements in the long end had already proved remarkably useful in anticipating changes in the Administration’s behaviour.

In August, however, the test became much more consequential.

The 30-year Treasury yield pushed through the area we had been watching.

And Treasury responded.

Beginning in September, liquidity-support buybacks in the 10-to-30-year part of the curve are being increased.

Now, let us be precise...

This is not QE. It is not yield-curve control (at least, not yet).

And the amounts involved are tiny relative to a Treasury market measured in tens of trillions of dollars.

But we think focusing on the amount risks missing the message.

The long bond produced a policy response.

That is what we wanted to know.

The TACO Zone was never about finding one magical yield at which Washington would suddenly panic. It was about identifying the point at which price would begin to influence behaviour.

Bessent and the Treasury have now given us the first important answer.

Their tolerance for restrictive conditions at the long end is not unlimited.

And, almost immediately, several of the assets we would expect to respond to that signal began doing exactly that.

Gold strengthened. Silver strengthened. Miners strengthened. Bitcoin strengthened. Ethereum strengthened.

Different assets. Different forms of scarcity. Different sensitivities.

But increasingly, the same message.

That also brings us back to another thesis we have been developing publicly for months:

The Dovish Shock.

From the beginning, our argument has been more nuanced than simply forecasting rate cuts.

Kevin Warsh does not need to become a dove.

He does not need to abandon his inflation-fighting credentials.

And he certainly does not need to engineer an emergency easing cycle.

The hurdle is much lower.

Policy merely needs to prove less restrictive than markets have already discounted.

Until recently, that thesis depended heavily on the Federal Reserve.

It no longer does.

Treasury has moved first.

And that is why the return of the Debasement Trade to our Leaderboard deserves more attention than another rally in precious metals.

The higher-beta parts of the complex are now doing particularly well. Junior miners have pushed back towards the top of the table. Altius is again approaching a triple. Silver has recovered strongly. Metals and mining have re-entered the Top 10.

The market appears to be travelling further out along the scarcity curve.

Crypto may be next.

It still does not appear in our Leaderboard.

We find that rather interesting.

A substantial part of the original asymmetry in gold, silver and miners has already been realised. Some of those positions have doubled. Some have been partially monetised. The trade has paid us handsomely.

Crypto is in a very different place.

The correction left technical damage behind. Sentiment remains far less enthusiastic. Bitcoin and Ethereum are only now beginning to repair their longer-term structures.

Yet the backdrop surrounding them is improving.

If Warsh eventually follows the Treasury... not by capitulating, but simply by proving less hawkish than the market expects... then some of our crypto exposures sit directly in the transmission channel.

And the next Leaderboard could look very different.

Our broader view on gold and crypto has always been that investors are asking the wrong question when they insist on choosing between them.

They represent different forms of scarcity.

They perform different jobs.

And in the right monetary environment, there is no reason both cannot work.

Meanwhile, just beneath the Top 10, something else is happening.

And it has almost nothing to do with Bessent, Treasury buybacks or the price of gold.

Our biotechnology theses are beginning to come alive.

Regular readers will know that this is not a new fascination.

Earlier this summer, we argued that one of the most interesting consequences of cheaper intelligence could emerge in industries where the cost of getting things wrong remains enormous.

Biotechnology fits that description unusually well.

Drug development is expensive. Failure rates are high. Clinical evidence is difficult to generate. Regulatory approval cannot be automated away. And modest improvements in the probability of success can create enormous economic value.

That was the premise behind our broader work on what we called The Intelligence Dividend.

We subsequently took that framework into two very different company-level investment cases.

One centred on a scientific discovery engine built around scarce biological evidence.

The other on a capital-light royalty model capable of participating in successful medicines without bearing the full economics of discovering them.

We are now making both of those pieces (originally published for paying subscribers) freely available as part of a broader release of selected research from behind the paywall:

Neither position has yet earned a place in the Top 10.

They are not alone.

And that is perhaps the most useful way to think about the Leaderboard.

It shows what has already worked.

It does not show the much larger group competing to get there.

Today, precious metals and real assets are once again dominating the visible table. Crypto may become the next beneficiary if the policy signal broadens. Biotechnology is beginning to build momentum for completely different reasons. And elsewhere across our product roster, several other positions are quietly moving closer to the threshold.

That is exactly what we want.

Not ten versions of the same macro bet.

But different sources of asymmetry, across different asset classes and different investment theses, capable of coming alive at different moments.

The public Leaderboard shows the winners once they have earned their place.

Subscribers get to see the race before the names appear on the board.

And, more importantly, they get to understand why we entered it in the first place.

Good Investing!

- Vasco

Important Disclosure

This article was produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It contains general investment research and market commentary and may include investment recommendations or information suggesting an investment strategy within the meaning of Regulation (EU) No 596/2014 on market abuse and Commission Delegated Regulation (EU) 2016/958. VMF Research’s activity as a producer of investment recommendations is subject to the applicable European market-abuse framework and to supervision by the Comissão do Mercado de Valores Mobiliários (CMVM) in Portugal.

This article is not personalised investment advice and has not been prepared by reference to any reader’s investment objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. References to previous recommendations, Model Portfolio positions, market scenarios or possible future Leaderboard entrants should be understood in the context of the research and investment horizons described in the relevant publications. Views may change as market conditions and evidence evolve and are reviewed through VMF Research’s monthly, weekly or ad hoc publications.

Research cut-off: 2 September 2026, following the close of the relevant markets.
First dissemination: 3 September 2026.

The Leaderboard is a selective presentation of VMF Research’s ten highest-returning recommendations across its paid publications and is not a complete record of its recommendations or Model Portfolio performance. Returns are measured from the original published reference price to the applicable exit price for closed positions or the relevant market price for positions that remain wholly or partially open. Unless otherwise stated, reported returns include recorded cash distributions and exclude taxes, transaction costs, custody charges, bid–ask spreads and foreign-exchange effects. The status column identifies positions that remain open, have been closed or have been partially realised. Displayed returns do not represent returns earned by clients or subscribers and should not be interpreted as the performance of an investable fund or managed account.

VMF Research’s Model Portfolios are illustrative research portfolios. They do not represent client assets, personalised mandates or transactions executed by VMF Research. References to the TACO Zone, the Dovish Shock, Treasury or Federal Reserve policy, precious metals, crypto assets, biotechnology and potential future market outcomes reflect research hypotheses, interpretations and forward-looking judgments that may not materialise. Linked articles were prepared at their respective original research dates and should be read together with the methodologies, assumptions, principal risks and disclosures contained in those publications.

Disclosure of interests: as at the research cut-off, legal entities controlled by the author hold long positions in Altius Minerals Corporation and the ARK Innovation ETF and hold long economic exposure to Bitcoin, Ether and Solana through listed exchange-traded products. These interests may create actual, potential or perceived conflicts and should be considered when evaluating the analysis. Interests may change after the stated disclosure date. Neither VMF Research nor the author received compensation from any issuer discussed in connection with the preparation of this article, and no issuer reviewed, approved or amended its investment conclusions before first dissemination.

Equities, precious-metals and mining securities, biotechnology investments, innovation-focused ETFs and crypto assets can be highly volatile and may result in the loss of some or all invested capital. Past performance, Model Portfolio performance and previously successful recommendations are not reliable indicators of future results. Readers remain responsible for conducting their own analysis and, where appropriate, should consult an authorised financial intermediary or adviser before making any investment decision.

More than usual, in fact.

But there was a reason for it. One piece of research kept opening the door to the next, and by the end of the month we had travelled from Chinese artificial intelligence to biotechnology, from biotechnology to two new stock recommendations, from China back into portfolio construction, and finally from digital abundance to one of the oldest physical bottlenecks in the world.

The journey was not planned in advance. That is partly what made it interesting.

It began with a question we have been pursuing for several months:

What happens to investment returns when intelligence becomes dramatically cheaper?

July gave us some answers.

China changed the economics

The natural place to begin is The China AI Reset.

The conventional AI debate remains obsessed with which country owns the most powerful model. We became more interested in a different question.

What if China does not need the best model in the world?

What if it only needs intelligence that is capable enough, improving quickly enough and cheap enough to perform an increasing share of economically useful work?

Chinese developers have been operating with less capital and more restricted access to advanced semiconductors than their American counterparts. Instead of stopping progress, those constraints appear to have encouraged a different kind of competition, centred on efficiency and cost.

That has important consequences.

If capable intelligence becomes cheaper, some of the scarcity premiums that defined the first AI boom should eventually come under pressure. At the same time, the businesses buying that intelligence receive a more powerful input at a lower price.

This is the essence of the Abundance Shock.

Value does not disappear when something becomes abundant. It migrates towards whatever remains scarce.

That led us directly into healthcare.

From chips to cures

We had already seen something happening in the market.

In Gold Steps Back. Biotech Steps Forward, our public Leaderboard showed Grail still sitting at the top of our realised winners, Roivant above 200%, and SBIO entering the leading group after a powerful advance.

Soon afterwards, we published The AI Trade May Be Beginning to Migrate From Chips to Cures.

Healthcare had been cheap for some time. We first made that case in October 2025. What July brought was something valuation cannot provide on its own: evidence that the market was beginning to care.

Earnings expectations were improving. Relative momentum was turning. Biotechnology had broken out. Large pharmaceutical companies were facing an increasingly important patent cliff, while mergers and acquisitions were accelerating.

Then AI added another dimension.

That became The Intelligence Dividend.

Drug discovery remains one of the most expensive exercises in trial and error in the economy. Choosing the wrong target, molecule or patient population can destroy years of work and enormous amounts of capital.

AI does not need to discover miracle drugs on demand to transform those economics. A modest improvement in target selection, trial design or the ability to abandon bad programmes earlier could create substantial value.

The cost of intelligence is falling while the cost of failure in biology remains extraordinarily high.

That gap caught our attention.

By the time we published When Cheap Starts to Lead, the investment case had moved beyond valuation.

Healthcare was no longer simply inexpensive. Parts of it were beginning to outperform.

That was when Tier Two went looking for companies.

We bought two very different businesses

July’s VMF’s Security Selection added two new positions to the Quality Model Portfolio.

The first was Regeneron, which we explored publicly in The Discovery Machine.

Regeneron has spent years building proprietary biological evidence through millions of sequenced samples linked with clinical and molecular information.

That becomes more interesting as AI gets cheaper.

If excellent algorithms become widely available, the algorithm itself becomes less scarce. The biological data on which it can work does not.

Regeneron therefore offered us a way to own something we think may become increasingly valuable in the next stage of AI: proprietary evidence.

Halozyme offered almost the opposite business model.

In The Royalty Engine, we explained why.

Regeneron must discover medicines. Halozyme can prosper when somebody else already has.

Its delivery technology is licensed to pharmaceutical companies, allowing Halozyme to participate through fees and royalties while its partners carry most of the development and commercial burden.

It is an economic architecture we know well. Altius uses royalties to participate in mines and power projects financed by other companies. Halozyme applies a comparable logic to biotechnology.

So we bought both.

Regeneron gave us the discovery machine.

Halozyme gave us the tollbooth.

And the early results have been encouraging. From the reference prices used when the recommendations entered the paid Model Portfolio on 17 July through 28 August, Regeneron had advanced approximately 17% and Halozyme approximately 35%.

Five weeks proves very little about a long term investment thesis, but it is difficult to complain when the evidence starts arriving quickly.

Stronger conviction did not mean buying more

The China research kept improving during the month.

New model releases added further evidence that Chinese developers were not simply competing by offering inferior intelligence cheaply. They were moving closer to the frontier while preserving attractive economics.

That became The Boom Changes Owners.

There was an obvious portfolio response.

Buy more China.

We didn’t.

The Thesis Strengthens. The Weight Does Not. explains the reason.

Alpha Tier already owned , , and , while Alibaba also represented a meaningful underlying exposure inside the ETFs themselves.

A stronger thesis did not change that arithmetic.

There is a difference between becoming more confident that something should be owned and concluding that it should become a larger position. The latter requires the next dollar of capital to offer a better expected return than every alternative available to the portfolio.

We did not think China had crossed that threshold.

The research produced one of the lines that best captures our approach:

Fundamentals tell us what deserves to be owned. Price tells us when conviction deserves more capital.

So we kept the China exposure where it was and followed the scarcity argument somewhere else.

The digital boom meets a physical constraint

Cheap intelligence should encourage more intelligence consumption.

That means more models, more agents, more inference and more automated workflows. Each individual task may become cheaper while the total number of tasks expands enormously.

Eventually, all of that software requires electricity.

And electricity does not scale like software.

That became The Fuel Beneath the Boom.

The uranium thesis was already attractive before AI entered the equation. The existing nuclear fleet consumes more uranium than the world mines each year. New reactors, restarts and life extensions are increasing future demand, while governments increasingly care about the security of the fuel supply itself.

India is contracting uranium for generating capacity that has not yet been built. China has been securing long term supply. Even producer nations are considering strategic inventories.

We increasingly think uranium is beginning to behave less like an ordinary commodity and more like a strategic reserve asset.

Yet while the fundamental argument improved, uranium equities corrected violently.

had fallen almost 45%.

Unlike China, this was an opportunity where the portfolio did not already carry substantial concentration. The correction restored the asymmetry we wanted.

Alpha Tier initiated the position.

By 28 August, URNM was approximately 15% above our reference price.

Again, early days. But July’s three new recommendations, , and , had all begun working.

That is what made July particularly satisfying. The research was not only prolific. It was actionable.

And then came the TACO Zone

Running alongside all of this was the macro hypothesis we introduced in June: the market remained vulnerable to a Dovish Shock.

July did its best to disprove it. War returned to the Middle East, oil spiked, Kevin Warsh remained hawkish and Gold and Silver both broke technical support.

Yet other parts of the evidence refused to cooperate with the bearish story.

Bitcoin held. Oil struggled to sustain its geopolitical advance. The market remained heavily positioned for tightening, but the probability distribution began moving slightly away from the most hawkish outcome.

Then long term Treasury yields moved into what we called the TACO Zone.

We explained the idea in The Treasury Yield Signals That Predicted Trump’s Ceasefire.

At sufficiently high long term yields, the pressure spreads into mortgages, equities, fiscal arithmetic and the broader political economy. That creates incentives for Washington to reduce whatever part of the pressure it can influence.

The ceasefire provided the first validation.

Scott Bessent’s subsequent focus on reducing pressure at the long end has made the framework more interesting still.

The Treasury Secretary appears to understand the problem.

Kevin Warsh is not there yet.

We think he will get there in time.

The Dovish Shock was never built around the idea that Warsh suddenly becomes a dove. It requires something much less dramatic. A market priced for further tightening merely has to discover that the tightening it expects may never arrive.

If Bessent has already understood where the pain begins, we suspect the Federal Reserve will eventually have to confront the same reality.

Stay tuned.

Your July reading map

If you want to retrace everything we published publicly from July’s work, this is the order I would use:

1. Gold Steps Back. Biotech Steps Forward

2. The AI Trade May Be Beginning to Migrate From Chips to Cures

3. The China AI Reset

4. The Intelligence Dividend

5. When Cheap Starts to Lead

6. The Discovery Machine

7. The Royalty Engine

8. The Boom Changes Owners

9. The Thesis Strengthens. The Weight Does Not.

10. The Fuel Beneath the Boom

11. The Treasury Yield Signals That Predicted Trump’s Ceasefire

July produced a lot of research because the same idea kept travelling.

Cheaper intelligence led us towards proprietary biology. Better healthcare evidence led us into two individual businesses. Stronger China fundamentals tested our willingness to concentrate further. Digital abundance led us back towards scarce electricity and uranium. Rising Treasury yields eventually brought monetary policy and political incentives back into the picture.

The links above give free readers almost the entire intellectual trail.

What they do not fully reproduce is the moment when research becomes a portfolio decision.

That is where VMF Research’s paid publications begin.

And in July, those decisions have already started paying us.

Important Disclosure

This article contains general investment research and market commentary produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It is provided for informational purposes only and does not constitute personalised investment advice, portfolio management, or a solicitation to buy, sell, subscribe for or hold any financial instrument, security, cryptoasset, commodity exposure, fund, ETF, ETP or other investment product.

The article summarises research published by VMF Research during July 2026 across VMF’s Strategic Asset Allocation, VMF’s Security Selection and Alpha Tier, together with related excerpts subsequently published on VMF’s Market View. References to securities, portfolio allocations and investment decisions are intended to explain VMF Research’s research process and should not be treated as a complete investment thesis or as a substitute for the underlying paid publications, which contain the relevant assumptions, risks, valuation work, sizing considerations and monitoring criteria.

References to the subsequent performance of securities discussed in this article, including Regeneron Pharmaceuticals, Halozyme Therapeutics and the Sprott Uranium Miners ETF, are calculated from the published Model Portfolio reference prices and do not represent client results or transactions executed by VMF Research. Model Portfolios are illustrative research portfolios and do not represent managed accounts or client assets. Reported performance excludes, unless otherwise stated, taxes, transaction costs, custody charges, bid ask spreads and foreign exchange effects. Past performance and Model Portfolio performance are not reliable indicators of future results.

The views expressed are stated as of the publication date and may change without notice as prices, fundamentals, policy expectations or other evidence evolve. Forward-looking statements and hypotheses, including the Dovish Shock and TACO Zone frameworks, are inherently uncertain and may not materialise. All investments involve risk, including the possible loss of capital.

Disclosure of interests: legal entities controlled by the author hold long positions in the KraneShares CSI China Internet ETF and Altius Minerals Corporation, both of which are referenced in this article. Legal entities controlled by the author also hold listed ETPs providing economic exposure to Bitcoin, Ether and Solana. These interests may create actual, potential or perceived conflicts and should be considered when evaluating the analysis. Neither VMF Research nor the author received compensation from any issuer discussed in this article in connection with its preparation.

Readers should conduct their own due diligence and, where appropriate, consult an authorised financial intermediary or other suitably qualified professional before making any investment decision.

Four months ago, we published an article with a deliberately provocative premise:

Artificial intelligence is beginning to compress the cost of cognition itself.

That was The Abundance Shock.

The argument was never that everything was about to become cheap. Quite the opposite. We suspected that as useful intelligence became easier to reproduce, the economic value of the things that remained difficult to reproduce would rise: proprietary data, trusted distribution, intellectual property, embedded workflows, electricity, infrastructure and the physical bottlenecks required to make the whole system work.

Much of the research we have published since then has been an attempt to follow that idea downstream.

The early AI boom rewarded scarcity. Advanced processors, high-bandwidth memory, data-centre capacity and power were difficult to expand quickly, and the companies controlling those bottlenecks captured the economics.

Then the evidence began to shift.

In The China AI Reset, the important development was not another record-setting frontier model. It was evidence that increasingly capable intelligence could be delivered at radically lower prices.

Most businesses do not need the smartest system in existence. They need one capable of completing a valuable task reliably, securely and cheaply enough to leave room for an economic return. That is a much larger market.

From there, the thesis naturally moved towards the application layer.

If intelligence gets cheaper, value begins to migrate towards companies that already control customers, proprietary information, distribution and workflows.

The Boom Changes Owners pushed that argument further.

The Intelligence Dividend, The Discovery Machine and The Royalty Engine followed it into healthcare and biotechnology, where better intelligence can be especially valuable because the cost of a bad decision is so high.

And The Fuel Beneath the Boom brought us back to the physical world. Intelligence may be getting cheaper, but producing and deploying it requires staggering amounts of electricity, compute and infrastructure.

That is why we have never viewed abundance and scarcity as competing investment narratives.

They increasingly feed one another.

Which brings us to Part II.

Until now, most of our work on the Abundance Shock has asked an investment question:

Where does value migrate when intelligence gets cheaper?

August’s VMF’s Strategic Asset Allocation asks a much bigger one.

What happens to the economy if it gets cheap enough?

Human cognitive time sits inside almost every business. Companies pay people to analyse information, write software, review contracts, forecast demand, communicate with customers, conduct research, organise processes and make decisions. Historically, increasing the amount of that work required more skilled labour and more hours.

Machine intelligence changes that constraint.

And the relevant price is not the cost of a token.

It is the cost of completing a useful piece of work.

As capability improves and that cost falls, tasks that were previously uneconomic to automate begin crossing into commercial viability. Companies stop using AI merely as an assistant and start redesigning entire workflows around what it can do.

That is where the Abundance Shock stops being just a technology thesis.

It becomes a macroeconomic one.

VMF Research has argued since inception that the post-2020 world would be characterised by higher and more volatile inflation because nominal demand increasingly collides with constrained productive capacity. Fiscal activism, deglobalisation, defence spending, energy security and reindustrialisation all reinforce that framework.

The Abundance Shock is the first force we have encountered that could seriously push against it from the supply side.

Cathie Wood frames that possibility at the macro level: productivity rising fast enough to push unit costs lower even while economic activity accelerates. Marc Andreessen approaches it through consumer surplus: enormous value may accrue to companies and consumers without appearing neatly in conventional revenue or inflation statistics. Vinod Khosla focuses on perhaps the most consequential transmission mechanism of all: making scarce expertise far more abundant across healthcare, education, law, engineering and other labour-intensive services.

That is a serious idea.

It is not yet a proven one.

The productivity data are encouraging but noisy. General-purpose technologies often require years of organisational redesign before their full benefits appear. Measurement may itself be struggling to keep pace. And cheaper intelligence could create so much new demand that spending on compute, electricity and infrastructure continues rising even while the cost of individual tasks falls.

So we are not abandoning the inflation framework that has guided VMF Research.

We are stress-testing it.

The same AI buildout currently adding to demand for electricity, semiconductors, capital and construction may eventually expand productive capacity enough to ease the inflation pressure it helped create.

That possibility also sits directly beside our Dovish Shock thesis.

If productivity improves, unit labour costs weaken and underlying inflation moderates without a collapse in growth, Kevin Warsh would face a very different economy from the one investors have spent months preparing for.

That is what makes The Abundance Shock – Part II so important to us.

The original article asked what becomes scarce when intelligence becomes abundant.

This one asks whether abundance can become large enough to affect the price level, the labour market and the Federal Reserve itself.

The excerpt below was first published for paid Tier One subscribers on 7 August 2026. It examines the arguments of Wood, Andreessen and Khosla, the early productivity evidence and the reasons we now regard this thesis as a serious challenge to the macro framework that has guided VMF Research since inception.

We started by asking where the value would go if intelligence became cheap.

Now the more important question may be:

What if intelligence becomes cheap enough to change the economy around it?

Good reading.

That productivity test is critical because the first stage of the AI cycle has been unmistakably inflationary.

Data centres require semiconductors, memory, cooling systems, land, construction workers and vast quantities of dependable electricity. The capital committed to building that infrastructure competes with defence, energy security, reindustrialisation and other Fourth Turning priorities for many of the same scarce resources.

That is the part of the cycle the market can already see. The harder question is what happens when the intelligence produced by that infrastructure begins spreading through the rest of the economy.

This is the Abundance Shock we have been writing about for some time: the possibility that increasingly capable and inexpensive machine intelligence transforms cognition from a scarce economic resource into something available in almost unlimited quantities.

Intelligence is embedded in nearly every commercial activity. Businesses must interpret data, write software, evaluate contracts, communicate with customers, forecast demand, perform research, organise processes and allocate capital. Historically, most of that work required human time... an expensive input that is difficult to scale and constrained by the number of hours available.

AI changes the economics of that input. It does not need to match the world’s best expert across every discipline to matter. It needs to perform a sufficiently valuable task, with adequate reliability, at a cost materially below the existing alternative.

That threshold is already moving rapidly.

Stanford’s AI Index found that the cost of obtaining roughly GPT-3.5-level performance fell more than 280-fold between late 2022 and late 2024.

Since then, the price war has intensified, particularly as increasingly capable Chinese models have entered the market at dramatically lower price points. The precise comparisons vary by capability, token mix and usage pattern, but the direction is unmistakable: increasingly capable machine intelligence is becoming cheaper at extraordinary speed.

The important metric, however, is not the price of a token. It is the cost of completing a unit of economically useful work. A cheaper model can still prove expensive if it requires repeated attempts, extensive supervision or costly correction. But as capability improves alongside price, activities that were previously too expensive to automate become commercially viable. The cost of intelligence falls, the number of economically rational applications expands and companies begin redesigning complete workflows around what the technology can do.

This is where the Abundance Shock begins to challenge our original inflation framework. That framework argued that money and nominal demand were expanding faster than the economy’s physical capacity to respond. The AI countercase is that productive capacity itself may now begin expanding faster than conventional models assume.

Cathie Wood has made the most explicit macroeconomic version of this argument. She believes the market is positioned for stagflation when it should instead be preparing for a productivity-led deflationary boom. In her framework, artificial intelligence, robotics, energy storage, blockchain technology and biological innovation reinforce one another, driving unit costs lower even as real economic activity accelerates. She has gone as far as to argue that inflation could turn negative, allowing Kevin Warsh to reduce rates into economic strength rather than in response to recession.

Some of Wood’s numerical forecasts are deliberately aggressive. Productivity growth of 4–6% would represent an historic departure from the experience of recent decades, while her preferred real-time inflation measures are not directly comparable with the official indices used by the Federal Reserve.

We should treat that outcome as a scenario, not as established fact. But the mechanism is compelling. When output per worker rises faster, wages can increase without generating the same pressure on unit labour costs. Companies can produce more while employing proportionately fewer additional resources. The productivity dividend can then be distributed through some combination of higher margins, greater investment, better compensation and lower prices.

approaches the same issue from the company and consumer level. His focus is consumer surplus: the value created when a product or service becomes vastly more useful or affordable without a proportionate increase in what the customer pays. Andreessen argues that much of AI’s ultimate economic value may accrue not to the model providers themselves, but to the businesses and consumers using inexpensive intelligence to accomplish tasks that were previously costly, slow or impossible.

This matters because conventional accounting may capture only a fraction of the benefit.

If an AI system allows a small company to perform research that once required a large analytical department, the economic gain is real even when no new revenue category appears in the national accounts. If software improves dramatically while its price remains unchanged, measured inflation may fail to capture the quality adjustment. If an AI assistant saves an individual several hours each week without charging for each hour saved, much of the value appears as time and consumer surplus rather than recorded output.

Vinod Khosla extends the argument into the part of the economy where it could become most consequential: services. The great disinflationary achievements of globalisation were concentrated largely in manufactured goods. Services remained dependent on local labour, professional qualifications and the limited supply of human expertise. Healthcare, education, law, finance and other knowledge-intensive activities therefore continued becoming more expensive even as electronics, clothing and many physical products became cheaper.

Khosla believes AI can begin reversing that pattern by making expert-level knowledge available at extremely low marginal cost. A doctor supported by specialised AI systems may serve more patients. A teacher assisted by an adaptive tutor may provide more individualised instruction. An engineer may test more designs. A research team may evaluate more hypotheses. A small business may gain capabilities that previously belonged only to a large institution.

The professional does not need to disappear for supply to expand.
The human bottleneck merely needs to become less binding.

Together, Wood, Andreessen and Khosla describe three layers of the same potential shock. Wood describes the macroeconomic outcome: higher productivity and lower inflation. Andreessen explains the economic mechanism: falling costs and expanding consumer surplus. Khosla identifies the crucial transmission channel: making scarce expertise abundant inside the service economy.

That combination could become powerfully disinflationary. It could even produce outright deflation in the products, services and business models most directly exposed to collapsing intelligence costs.

But we should remain precise.

A fall in the price of software, analysis or administrative work is not the same as a sustained decline in the general price level. Housing, energy, infrastructure, commodities and many forms of physical labour remain constrained. Fiscal spending may absorb part of the new productive capacity. Cheaper intelligence may create so much additional demand that total spending on AI, compute and electricity continues rising even as the cost of each task falls.

AI can therefore be deflationary at the task level, disinflationary inside individual businesses and still fail to produce outright economy-wide deflation.

The best current evidence lies in productivity and unit labour costs.

The preliminary second-quarter data released today show US non-farm business productivity increasing at a 1.4% annualised rate and 2.2% from a year earlier. Unit labour costs rose only 1.4% over the same four-quarter period. Since the end of 2019, productivity has grown at an annualised rate of 2.1%, compared with 1.5% during the preceding business cycle.

The Federal Reserve has noticed. Its July Monetary Policy Report described productivity growth as strong and concluded that current wage growth, when considered alongside productivity, was roughly consistent with 2% inflation over time. The Fed also judged that AI’s aggregate contribution remains modest so far, suggesting that much of the potential Intelligence Dividend may still lie ahead.

Encouraging does not mean conclusive.

Productivity is real output divided by hours worked, and both sides of that calculation are estimated and repeatedly revised. Today’s release revised first-quarter productivity higher by half a percentage point. Total factor productivity is even more elusive because it is not directly observed; it is the residual left after statisticians estimate the contributions of labour and capital.

AI may also remain hidden in the data for longer than investors expect. General-purpose technologies require companies to redesign processes, retrain workers, improve data and develop entirely new organisational structures. Those complementary investments impose costs before their benefits appear. Economists Erik Brynjolfsson, Daniel Rock and Chad Syverson describe this as the Productivity J-Curve: measured productivity can initially disappoint while organisations build the intangible capital needed to use the technology, then accelerate later as those investments begin paying off.

Measurement itself may also be falling behind the economy. The Bureau of Economic Analysis is developing new statistics for AI-related production and has found that treating internally created data as a capital asset would increase the estimated contribution of IT-related capital to US growth by roughly one-third between 2002 and 2024.

The absence of an obvious AI boom in today’s aggregate productivity figures would therefore not disprove the thesis.

But neither should every favourable productivity release be attributed to AI.

Business formation, capital deepening, labour-market composition, remote work and post-pandemic reorganisation may all be contributing. The signal will emerge gradually and probably unevenly across industries.

That is why we will monitor more than one number.

We will watch productivity over rolling four-quarter periods, unit labour costs, wage growth relative to output per hour, the price of completing real AI tasks, adoption inside commercial workflows and inflation across the service categories most exposed to machine intelligence. We will also examine where the benefit is going: towards lower consumer prices, higher corporate margins, increased wages or another wave of capital expenditure.

The Abundance Shock is not yet strong enough to overturn our original Market View.

The structural forces behind higher and more volatile inflation remain in place. The physical construction of the AI economy is reinforcing several of them. But the technology being built with those scarce resources may become the most powerful positive supply shock of our generation.

That possibility represents a genuine test of the framework that has guided VMF Research since its inception... and we will treat it accordingly.

We will monitor the evidence relentlessly, search actively for what could prove us wrong and watch these trends like hawks, pun very much intended.

Whether the Abundance Shock already justifies a change to the Model Portfolio is the question we address next.

You might also like reading Part I

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the August 2026 issue of VMF’s Strategic Asset Allocation. The research, views and Tier One Model Portfolio information are stated as of 7 August 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 7 August 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Tier One Model Portfolio is an illustrative research portfolio designed to communicate VMF Research’s strategic asset-allocation framework, Market View and investment conclusions. It does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines macroeconomic, technological, monetary-policy, labour-market and strategic asset-allocation research within a medium- to long-term framework. Statements concerning artificial intelligence, productivity, unit labour costs, consumer surplus, the future cost of economically useful intelligence, disinflation, deflation and the potential effects of technological adoption on economic growth and Federal Reserve policy are analytical judgments and conditional hypotheses rather than assurances.

References to Cathie Wood, Marc Andreessen and Vinod Khosla describe and interpret views relevant to the economic scenarios considered in the research. They should not be understood as implying any affiliation with, endorsement by or participation of those individuals or their organisations in VMF Research’s analysis.

The economic effects of artificial intelligence remain highly uncertain. Falling model or inference costs may not translate into equivalent reductions in the cost of completing economically useful work. Adoption may require substantial additional investment, organisational redesign, data infrastructure and human supervision. Productivity gains may arrive later or prove smaller than anticipated, while growing demand for semiconductors, electricity, data centres and other physical infrastructure may offset part of any disinflationary effect. Improvements in productivity may also be driven by factors unrelated to artificial intelligence, and official economic data are subject to estimation error and revision.

Neither VMF Research nor the author received compensation from any individual, issuer, fund sponsor or other entity covered in connection with the preparation of this research, and no covered entity reviewed, approved or amended its investment conclusions before first dissemination.

Past performance, Model Portfolio performance, historical economic relationships and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

Two weeks ago, the market was still asking how soon Kevin Warsh might raise rates.

We were asking a different question.

What if he did not have to?

On 7 August, when we published this section in VMF’s Strategic Asset Allocation, Warsh’s reputation was already doing much of the analytical work for investors. He had arrived at the Federal Reserve determined to restore credibility, repeatedly defended the 2% inflation target and deliberately reduced the forward guidance markets had grown accustomed to. Three FOMC members had just voted for an immediate rate increase.

The conclusion seemed straightforward: Warsh was a hawk, therefore the next meaningful move in policy was more likely to be tighter.

We thought that interpretation was too simple.

His rhetoric was uncompromising, but his behaviour was more nuanced. He had already supported holding rates through two meetings. More importantly, he had drawn a distinction between individual price shocks and a sustained broadening of underlying inflation. Tariffs, energy, geopolitical disruption and the enormous capital spending associated with AI could all lift particular prices without necessarily requiring the Federal Reserve to respond mechanically.

That distinction sat at the heart of our Dovish Shock thesis.

We were never forecasting an imminent cutting cycle. The more interesting possibility was much less dramatic: markets had become so conditioned to expect tighter policy that a transition from expected tightening to patience could be enough to change real yields, the dollar, liquidity conditions and cross-asset leadership.

Tier One had already been positioned with that possibility in mind. And over the past few days, the evidence has moved noticeably in our direction.

July consumer and producer inflation both came in softer than expected, while the labour data also weakened. By 14 August, Reuters was describing the new information as potentially setting the Fed up for an “extended do-nothing stance”, with traders abandoning much of their expectation for a September hike.

Then came weaker retail sales.

By 17 August, market pricing for a September increase had flipped towards a roughly 70% probability of no change. In Reuters’ latest economist poll, 90% expected the Fed to hold in September and nearly 80% expected no change through year-end. That is a meaningful shift from the environment in which this research was written.

It does not prove the Dovish Shock.

But the market is beginning to travel towards the scenario we described.

At the same time, another part of our August argument has become almost impossible to miss.

We called the area around 5.10–5.20% on the 30-year Treasury the TACO Zone because this was where an abstract argument about inflation and fiscal policy began imposing very real costs on mortgages, corporate financing, equity valuations and government borrowing. The level itself was never meant to be magical. The important point was the feedback mechanism: long-term yields could eventually constrain political choices in ways the White House could not simply talk away.

This week, the 30-year yield reached 5.34%, its highest level since 2007.

The Treasury responded by doubling its long-end bond buybacks to at least $4 billion per operation. Reuters noted that the move revealed the administration’s sensitivity to higher long-term borrowing costs and its willingness to intervene once those yields became sufficiently uncomfortable.

That is remarkably close to the mechanism we described on 7 August.

Warsh can control the overnight policy rate.

The administration can pressure the Fed.

Neither can casually ignore the long bond.

This is why the interaction between the hawk and the TACO Zone matters. Warsh’s credibility requires him to convince investors that 2% still means 2%. Yet the stronger that credibility becomes, the more freedom he may eventually have to respond to softer inflation and weaker activity without markets interpreting patience as surrender.

The recent data are beginning to test that possibility in real time.

There is still plenty that could reverse it. Inflation remains above target, oil remains elevated and several FOMC members continue to argue that tighter policy may ultimately be necessary. Markets still assign a meaningful probability to a rate increase later this year. Jackson Hole may move the debate again.

But that is precisely why revisiting this research now is useful.

What follows was first published to paid Tier One subscribers on 7 August 2026 at 9:00 p.m. Eastern Daylight Time, before the latest inflation releases, before the weaker retail data and before the Treasury was pushed into increasing its intervention at the long end.

We had been waiting for evidence that the market’s image of Warsh as a permanently hawkish Fed Chair was too rigid.

We are starting to get it.

And the bond market is simultaneously showing us why the cost of staying hawkish may be considerably higher than it looked only a few weeks ago.

Good reading.

The inflationary lens has become so dominant that it now shapes the interpretation of policy before policy is even made.

Kevin Warsh is the clearest example. The market sees a central banker determined to restore the Federal Reserve’s credibility, unwilling to tolerate another period of above-target inflation and prepared to reconsider many of the institution’s established practices.

It sees a hawk.

Warsh has certainly given investors reasons to reach that conclusion. In his first testimony to Congress as Chair, he described price stability as a central obligation of the Federal Reserve and insisted that persistent inflation must ultimately be owned by monetary policy. There would be no quiet migration towards a higher target and no attempt to explain away a generalised rise in prices through an endless succession of external shocks. The Federal Reserve would return inflation to 2%.

The language has been deliberately uncompromising.

His actions have been more nuanced.

Warsh has chaired two FOMC meetings since taking office. In June, the Committee unanimously maintained the federal funds rate at 3.50–3.75%. In July, Warsh again supported holding rates even though three members preferred an immediate quarter-point increase. This is not the record of a Chair rushing to validate his hawkish reputation through higher policy rates.

Nor has Warsh described monetary policy as a mechanical response to every uncomfortable inflation release. He has distinguished between increases in individual prices and a broader movement in underlying inflation. Energy, tariffs, geopolitical conflict and AI-related investment may all raise costs in particular sectors. The policy question is whether those increases spread into unrelated categories, alter expectations and become embedded across the economy.

His reaction function is consequently more symmetric than the market narrative suggests. With the labour market close to equilibrium, rising underlying inflation would make him more inclined to tighten. Falling underlying inflation would make him more inclined to loosen.

The target is fixed.
The route is not.

Warsh’s most consequential change may therefore concern not the current level of the federal funds rate, but the way the Federal Reserve communicates. He has reduced the forward guidance through which central banks had become accustomed to preparing investors for almost every policy decision. Markets should respond more directly to economic information rather than continually attempting to anticipate the Fed’s next carefully rehearsed sentence.

That approach has merit. A central bank that manages every movement in financial markets risks replacing price discovery with a feedback loop in which investors merely repeat the Fed’s own forecasts back to it. Less guidance may eventually produce more informative market prices.

But it also means that those prices must be taken seriously.

Long-term yields rose sharply after the July meeting even though the policy rate remained unchanged. The thirty-year Treasury yield moved above 5.20%, reaching its highest level since 2007, as investors questioned whether the Federal Reserve would act forcefully enough to restore inflation to its target. Reuters described the move as a test of Warsh’s credibility rather than a simple response to the FOMC decision itself.

Warsh wants the market to speak.
The long bond has taken him literally.

The chart above focuses on the part of the Treasury curve that increasingly matters for both monetary and political policy.

The thirty-year yield has established a sequence of higher lows and repeatedly tested the area between approximately 5.10% and 5.20%.

We have labelled that region the TACO Zone.

The name is intentionally irreverent.
The constraint is not.

A thirty-year yield above 5.20% raises long-term mortgage costs, increases the hurdle rate for corporate investment, weakens the valuation support provided by distant cash flows and raises the government’s prospective refinancing burden. It is where the abstract debate about inflation begins imposing an immediate cost on the economy and the political system.

The significance of that level has become especially visible during the war with Iran.

The conflict disrupted energy shipments through the Strait of Hormuz, lifted oil and petrol prices and forced investors to reassess the likelihood of further Federal Reserve tightening. As inflation risk rose, the long end sold off. In May, the thirty-year yield reached 5.20% for the first time since 2007, while markets moved from expecting several rate cuts before the war towards assigning a high probability to a rate increase.

At approximately the same time, President Trump began moderating his demands for immediate monetary easing. Having previously expressed disappointment at the prospect that Warsh might not cut rates quickly, he adopted a more patient tone and acknowledged that the Fed Chair would have to respond to the conditions confronting him. The timing does not prove that the bond market alone forced the rhetorical shift, but it showed how quickly political ambitions could collide with long-term borrowing costs.

The pattern subsequently repeated itself across the Iran conflict.

In June, Trump cancelled planned strikes against Iran only hours after threatening to hit the country “very hard”. Brent and WTI fell by almost 3% after the announcement, even though Iran had not confirmed the agreement Trump suggested was advancing. Reuters noted that the President had repeatedly announced that a deal was imminent, only to return to military threats when Tehran refused his terms.

The same sequence became visible again in early August. Trump withheld further action and claimed talks were under way, while Iran denied that negotiations had been scheduled. Brent fell approximately 7%, Treasury yields declined and equities rallied as markets priced another possible de-escalation. The thirty-year yield retreated after having reached 5.28%, its highest level since 2007.

This is the geopolitical version of the TACO trade.

The threat creates the market shock.
The market shock increases the domestic political cost.
The threat is softened before that cost becomes intolerable.

The TACO Zone should therefore not be understood as one magical number on one chart. It is a broader pressure point connecting the Iran war, oil prices, inflation expectations and long-term borrowing costs.

When escalation threatens the Strait of Hormuz, oil rises.
When oil rises, the inflation problem confronting Warsh becomes more difficult.
When the expected policy path and inflation risk rise, the long bond approaches or breaks through the 5.10–5.20% zone.

And when those yields begin tightening mortgages, corporate finance, equity valuations and government funding simultaneously, Trump’s room for manoeuvre narrows.

Retail investors have increasingly recognised the pattern. Reuters reported that “TACO” has moved into everyday market language, with traders buying declines created by Trump’s most aggressive announcements on the expectation that economic or market pressure will eventually force another retreat. The long bond has become one of the clearest measures of that pain threshold.

None of this means every change in Trump’s Iran policy is caused by Treasury yields. Military capacity, domestic approval, petrol prices, diplomacy and the actions of Iran and America’s regional partners all matter. Nor does it mean 5.20% will remain the relevant threshold indefinitely.

But the repeated interaction is increasingly difficult to ignore.

Trump wants lower policy rates, stronger growth, greater defence production, more domestic manufacturing and enormous investment in energy and artificial intelligence. Each objective may be individually defensible. Collectively, they require capital and risk placing further upward pressure on long-term yields.

The President can pressure the Federal Reserve.
He cannot order the thirty-year Treasury yield lower.

This brings us to what we believe the market may be misreading about Warsh. He assumed office under unusually difficult circumstances. Inflation had remained above target for more than five years. The President who appointed him had repeatedly demanded easier monetary policy. Investors therefore had legitimate reasons to question whether the new Chair would defend the mandate independently of the White House.

A Chair entering under those conditions could not establish credibility by sounding dovish. He had to demonstrate that 2% still meant 2%, that the Federal Reserve would not automatically accommodate fiscal or political pressure and that he was willing to disappoint the President who appointed him.

Our interpretation is that Warsh’s hawkish language is partly an exercise in rebuilding that institutional credibility. This is an inference rather than something he has explicitly acknowledged, but it is consistent with his emphasis on accountability, independence and the need to put the recent period of high inflation behind the institution.

Credibility, however, is not the same thing as permanent hawkishness.
It is what preserves the freedom to respond when the evidence changes.

A Chair who convinces markets that he will defend the inflation target may eventually be able to loosen policy without appearing to have abandoned it. A Chair whose independence remains in doubt cannot.

This is the central asymmetry behind our Dovish Shock1 thesis.

You might also like reading, from June’s VMF’s Strategic Asset Allocation, Fighting the Wrong War:

Warsh may need to sound more hawkish today precisely so that he can become more dovish later. Not because he changes the destination, but because the inflation confronting him changes.

The market has largely collapsed those two possibilities into one. It sees Warsh’s commitment to price stability and assumes a permanently restrictive policy path.

Yet his decisions, his treatment of supply shocks and his stated reaction function leave considerably more flexibility than that interpretation allows.

We are not arguing that rate cuts are imminent. We are arguing that the hurdle for a dovish repricing may be lower than investors believe.

The Federal Reserve does not need to cut aggressively for the Dovish Shock to matter. A shift from expected tightening towards an extended hold could be sufficient to alter real yields, the dollar, liquidity conditions and cross-asset leadership.

The most revealing evidence may be found outside the immediate rates debate. Warsh has established a dedicated task force to examine productivity and employment in an economy being transformed by general-purpose technologies, including artificial intelligence. The group is explicitly studying what those developments mean for productive capacity and for the Federal Reserve’s inflation and employment mandates. is one of its external leaders.

Warsh has also highlighted the extraordinary scale of AI-related capital expenditure. High-technology investment is growing rapidly, driven by data centres and the equipment and software required to operate them. But the Fed is not only examining the immediate demand for semiconductors, electricity and construction. It is asking what the technology could eventually mean for the supply side of the economy.

That question leads directly to the other side of this issue.

The market sees the power plants, data centres, semiconductors and financing required to construct artificial intelligence. Warsh is already asking what that investment may do to the economy’s future capacity. And if AI allows supply to expand faster, unit costs to fall and underlying inflation to moderate, his own reaction function tells us what follows.

The hawk becomes more inclined to loosen.

The market may believe Warsh’s credibility depends on remaining hawkish. We believe that credibility may eventually be what permits the dovish turn.

But for that thesis to work, the technology must deliver more than investment demand.

It must deliver productivity.

That is where we turn next.

1

We presented our Dovish Shock thesis in detail in June’s VMF’s Strategic Asset Allocation, Fighting the Wrong War. The argument was never that inflation had disappeared or that aggressive rate cuts were imminent. It was that markets had become so conditioned to a hawkish Federal Reserve that even a shift from expected tightening towards patience could trigger a meaningful repricing in real yields, liquidity conditions and cross-asset leadership.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It incorporates an excerpt from the August 2026 issue of VMF’s Strategic Asset Allocation, together with introductory commentary on developments that occurred after the original research cut-off. The underlying Tier One research and Model Portfolio information are stated as of 7 August 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 7 August 2026 at 9:00 p.m. Eastern Daylight Time. Subsequent economic, policy and market developments referred to in the introduction are identified by their relevant dates.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Tier One Model Portfolio is an illustrative research portfolio designed to communicate VMF Research’s strategic asset-allocation framework, Market View and investment conclusions. It does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines macroeconomic, monetary-policy, fixed-income, geopolitical and market research within a medium- to long-term framework. Statements concerning Kevin Warsh’s reaction function, the Dovish Shock thesis, inflation, Federal Reserve policy, Treasury yields, oil prices, geopolitical developments and the TACO Zone are analytical judgments and conditional interpretations rather than assurances. References to the TACO Zone describe an observed market pressure point and analytical framework; they should not be interpreted as implying that any particular Treasury yield constitutes a permanent threshold or will reliably produce a specific policy, political or market response.

Economic releases, market-implied probabilities and policy expectations can change rapidly and may be revised. A period of softer inflation or weaker activity may not lead to easier monetary policy, lower real yields or the cross-asset outcomes contemplated in the analysis. Inflation may remain persistent, energy or geopolitical shocks may intensify, Federal Reserve policy may prove more restrictive than expected and historical relationships between interest rates, financial conditions and asset prices may fail.

Neither VMF Research nor the author received compensation from any issuer, financial institution, public official or other entity covered in connection with the preparation of this research, and no covered entity reviewed, approved or amended its investment conclusions before first dissemination.

Past performance, historical market relationships, Model Portfolio performance and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

Lacy Hunt changing his mind deserves more attention than another hot CPI print.

For decades, Hunt was one of the clearest advocates of the secular deflation case. His framework was simple but powerful: excessive debt eventually drags on economic activity. More income is diverted towards servicing past borrowing, each additional dollar of credit produces less growth, money velocity falls and inflation weakens.

For a long time, the evidence was on his side.

After 2008, central-bank balance sheets exploded, government debt kept climbing and economists repeatedly warned that inflation was around the corner. Yet consumer-price pressure remained subdued, growth disappointed and long-dated Treasury yields continued their forty-year decline.

Hunt understood something many others did not. Creating liquidity was not the same as creating spending.

What makes his latest work so interesting is that he has not suddenly decided debt no longer matters. He has gone back and asked why the old relationship between debt, growth and inflation worked so reliably for so long.

His answer will sound familiar to regular readers of VMF Research.

The debt supercycle coincided with an extraordinarily benign supply environment. China was integrating into the global economy. Hundreds of millions of workers joined the global labour pool. Production moved towards lower-cost locations. Supply chains became ruthlessly efficient. Energy was cheap, capital plentiful and globalisation continuously expanded the economy’s ability to meet demand.

Debt restrained spending while supply kept getting easier to expand.

That combination helped produce the disinflationary world Hunt spent decades describing.

It is no longer the world we live in.
Today, commercial efficiency increasingly takes second place to national security. Governments want domestic semiconductor capacity, stronger electricity grids, more defence production, strategic inventories and less dependence on geopolitical rivals. Capital is being demanded simultaneously by defence, energy, AI, reindustrialisation and infrastructure.

These are exactly the forces behind the higher-and-more-volatile inflation regime we have written about since VMF Research began publishing in May 2024 (and, frankly, the framework through which we had already been looking at markets well before then).

Hunt has now travelled a long way in the same direction.

He believes the long-run US inflation range may have shifted materially higher and that interest rates are likely to become much more volatile than during the great disinflation. For someone whose name became almost synonymous with debt-driven deflation, that is a meaningful change of view.

And this is where things become more interesting for us.

There is little value in spending the next five years repeating a thesis the rest of the market has finally learned.

When we launched VMF Research, the idea that inflation, interest rates and policy volatility were moving into a structurally higher range still sat well outside the comfortable post-2008 consensus. Today, fiscal dominance, deglobalisation, industrial policy, energy security and capital scarcity are standard macro vocabulary.

Even Lacy Hunt is now looking through much the same lens.

So what are we supposed to see that the market does not?

His own work provides part of the answer.

Hunt may have revised his secular view, but he has not discarded the cyclical mechanisms that made him such an effective deflationist. An oil shock can lift headline inflation and long-term yields, then weaken household purchasing power, corporate margins and employment badly enough to suppress demand. Monetary restraint still matters. Credit stress still matters. Leverage still matters.

And there is another force worth taking seriously: productivity.

Artificial intelligence is currently consuming enormous amounts of capital, electricity, semiconductors and construction capacity. From that angle, it fits perfectly inside the inflationary regime.

But we are not building data centres for the pleasure of owning data centres.

We are building them because the intelligence they produce may eventually allow businesses to do more with less labour, less time and lower marginal cost.

If that begins to happen at scale, the next disinflationary impulse could emerge from inside the very investment boom that currently looks inflationary.

That does not require us to abandon the post-2020 framework. It requires us to do what good investors should do when a thesis works: raise the standard of evidence and look harder for the next thing the market is underestimating.

The excerpt below was first published for paid subscribers on 7 August 2026. It examines what made Hunt’s old deflationary framework work, why he now believes the environment has changed and why his remaining instincts may prove particularly useful now that the secular inflation thesis has become mainstream.
Lacy Hunt has moved towards our side of the argument.

That is gratifying.

It is also a good reason to start looking somewhere else.

Good reading.

Few economists have defended the secular deflationary case with greater consistency, intellectual rigour or longevity than Lacy Hunt.

For decades, Hunt argued that excessive debt would ultimately suppress rather than stimulate economic activity.

The mechanism was straightforward.

Debt allows expenditure to be brought forward. But the resulting obligations must eventually be serviced from future income. As debt accumulates, a larger share of household, corporate and government resources is redirected towards interest and principal payments rather than current consumption or productive investment.

Each additional unit of borrowing generates progressively less economic activity.

Growth slows.
Money circulates less rapidly.
Inflation weakens.
Long-term interest rates decline.

That framework placed Hunt far outside the dominant inflationary interpretation of monetary policy after the Global Financial Crisis. Many analysts looked at the extraordinary expansion of central-bank balance sheets and concluded that runaway consumer-price inflation was inevitable.

Hunt saw something different.

He observed that liquidity was accumulating inside a heavily indebted financial system whose capacity and willingness to spend were deteriorating. The Federal Reserve could increase the quantity of money, but it could not compel households, companies and banks to deploy it productively.

Money existed.
Velocity collapsed.
The real economy remained subdued.

For much of the period between 2008 and 2020, Hunt was right.

The Federal Reserve expanded its balance sheet by several trillion dollars. Government debt continued rising. Yet inflation repeatedly undershot expectations, economic growth remained modest and long-term Treasury yields continued the secular decline that had begun during the Volcker era.

The thirty-year Treasury yield fell from double-digit levels in the early 1980s to below 1% during the Covid Crash.

The consensus repeatedly waited for inflation.

Hunt repeatedly explained why it failed to arrive.

His latest work does not discard that framework. It asks a more fundamental question: what made it work for so long?

Hunt now places much greater weight on the unusually favourable supply environment that accompanied the globalisation era. China’s integration into the world economy, the arrival of hundreds of millions of lower-cost workers, increasingly efficient supply chains, abundant capital and cheap energy created one of the largest positive supply shocks in modern history.

Productive capacity expanded while debt servicing restrained demand. At the same time, money velocity declined sharply, allowing enormous increases in debt and liquidity to be absorbed without producing sustained consumer-price inflation.

Debt was disinflationary, but it did not operate in isolation. It operated inside a world in which supply was unusually elastic.

That world is disappearing.

Strategic rivalry is replacing commercial integration. Tariffs, friendshoring and domestic semiconductor production are replacing the lowest-cost-producer model.

Supply chains are being redesigned around resilience and redundancy. Labour-force growth is slowing, while defence, energy, grid infrastructure, reindustrialisation and artificial intelligence are competing for increasingly scarce capital and skilled workers.

In Hunt’s revised framework, liquidity can no longer be assumed to disappear harmlessly into a constantly expanding global production system. It may instead collide with labour shortages, insufficient energy infrastructure, lower economies of scale and physical bottlenecks. Rising debt may still weaken long-term growth, but it need not remain disinflationary if policymakers continue creating nominal demand faster than productive capacity can respond.

That is a significant intellectual evolution.

Hunt now believes that the long-run US inflation range may be migrating from approximately 1.5–3.5% towards 3.5–4.5%, with a meaningful risk of episodes above 5%. He also expects interest rates to become more volatile as higher inflation expectations, greater demand for real capital and a less dependable Treasury term premium replace the forces that drove yields steadily lower between 1990 and 2020.

Yet his first-quarter analysis of the latest oil shock shows that he has not abandoned the cyclical forces at the heart of his earlier work.

An energy shock initially appears inflationary because it raises the cost of almost everything that must be produced, heated or transported. Petrol becomes more expensive. Utility bills rise. Distribution costs increase. Headline inflation responds quickly, and long-term yields may rise before the damage to economic activity becomes visible.

But Hunt argues that an oil shock striking an already leveraged and financially vulnerable economy carries a second, more persistent consequence.

Higher energy bills reduce household purchasing power. Businesses attempt to pass those costs to customers, but weak demand limits their ability to do so. Margins compress. Investment is postponed. Hiring slows. Lay-offs increase. Delinquencies and bankruptcies rise as already fragile balance sheets absorb another blow.

The initial inflation surge can therefore be temporary even when the economic damage is not.

As households reduce spending and companies lose pricing power, the slowdown eventually suppresses the very inflation the shock created. The economy experiences what Hunt considers more accurately described as a cost-push recession rather than simple stagflation: prices rise first, but output and employment subsequently fall. Inflation fades because demand has been weakened, not because productive capacity has improved.

The oil shock is therefore less the origin of the problem than the catalyst that exposes it. Hunt finds a similar pattern in earlier episodes, when energy-price surges arrived in economies already burdened by slowing growth, financial stress or excessive leverage.

The visible shock came from oil... the lasting damage emerged through incomes, profits, employment and credit.

This distinction also explains why Hunt does not advocate an aggressive monetary response to every temporary increase in inflation. Tightening into a supply shock cannot produce more energy. It can only weaken demand further. Easing aggressively risks allowing the initial price increase to spread into expectations and broader inflation.

His preferred response is restrained: tolerate the temporary impulse, preserve credibility and allow the economy’s own slowdown to prevent it from becoming permanent.

Hunt therefore continues to see important routes towards disinflation. Recession can destroy demand. Monetary restraint can slow money and credit. Financial stress can reduce spending. A favourable technological shock can expand productive capacity.

What has changed is the environment surrounding those cyclical forces.

Disinflation is no longer the automatic destination of every increase in debt. It has become one possible movement inside a structurally higher and more volatile inflation regime.

Artificial intelligence occupies an especially important place in that debate. Hunt recognises that AI may eventually generate meaningful productivity gains. Better software, automated processes and more efficient decision-making could allow the economy to produce more from the same labour and capital.

His concern is about sequencing.

The productivity dividend may arrive later, while the capital expenditure required to build it is arriving now.

Data centres require semiconductors, cooling systems, transmission infrastructure and enormous quantities of dependable electricity. Those investments are being made while the grid is already constrained and while defence, domestic manufacturing and energy infrastructure are drawing upon many of the same resources. Hunt therefore believes that AI’s near-term effect may be more inflationary than disinflationary: the economy must first finance and construct an extraordinarily resource-intensive system before it can harvest its productivity gains.

That is a serious argument, and we do not dismiss it.

We deeply respect Hunt’s work, the analytical consistency with which he defended his earlier views and the intellectual honesty required to revise them when the underlying conditions changed. His historic deflationary framework explained the post-2008 world far better than the repeated predictions of runaway inflation made by many of his contemporaries.

His new framework is also remarkably close to the Market View that has shaped VMF Research since its inception.

We have argued that the combination of fiscal activism, deglobalisation, capital scarcity, geopolitical fragmentation and constrained productive capacity would leave inflation and interest rates higher and more volatile.

That view informed our large Alternatives Book, our exposure to natural resources and our deliberate underweight to conventional fixed income.

It has benefited the Model Portfolio materially.

Hunt’s turn does not tell us that the view is wrong.

It tells us how far it has travelled...

When one of the modern era’s most committed and respected deflationists identifies inflation as the greater structural risk, the inflation thesis can no longer be described as marginal or contrarian. It has become the central macroeconomic framework through which investors interpret the world.

Geopolitical conflict is viewed first through energy prices.
Defence expenditure through capital scarcity.
Fiscal policy through Treasury supply.
Industrial policy through labour and commodity constraints.
Strong growth through the prospect of tighter monetary policy.

Even AI is initially interpreted through the electricity, semiconductors and investment it consumes rather than the costs it may eventually eliminate.
The inflationary lens is not necessarily wrong.

But it has become so dominant that nearly every new development is now made to confirm it.

And that is where we turn next.

You might also like reading:

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the August 2026 issue of VMF’s Strategic Asset Allocation. The research, views and Tier One Model Portfolio information are stated as of 7 August 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 7 August 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Tier One Model Portfolio is an illustrative research portfolio designed to communicate VMF Research’s strategic asset-allocation framework, Market View and investment conclusions. It does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines macroeconomic, monetary-policy, fixed-income, geopolitical, energy-market, technological and strategic asset-allocation research within a medium- to long-term framework. Statements concerning debt, money velocity, inflation regimes, interest rates, Treasury yields, oil shocks, monetary-policy transmission, artificial intelligence and productivity are analytical judgments and conditional hypotheses rather than assurances. References to Lacy Hunt describe and interpret views discussed in the underlying research and should not be understood as implying any affiliation with, endorsement by or participation of Mr Hunt in VMF Research’s analysis.

Investments and asset classes discussed within VMF’s Strategic Asset Allocation may involve substantial risks, including changes in inflation and interest rates, duration and credit risk, commodity-price volatility, geopolitical disruption, policy error, currency movements, liquidity constraints, technological disruption and the possible loss of capital. Historical relationships between debt, inflation, economic growth and interest rates may change. Oil shocks may produce different economic or market outcomes from previous episodes, and potential productivity gains from artificial intelligence may arrive later, prove smaller or affect inflation differently than anticipated. Macroeconomic, valuation and technical frameworks may fail.

Neither VMF Research nor the author received compensation from any issuer, fund sponsor, individual or other entity covered in connection with the preparation of this research, and no covered entity reviewed, approved or amended its investment conclusions before first dissemination.

Past performance, Model Portfolio performance, historical market relationships and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

The most dangerous moment for a good investment thesis is not when the market disagrees with you. It is when the market finally agrees.

This is a nice problem to have. But it is still a problem.

When VMF Research began publishing in May 2024, our central macro view was decidedly outside the mainstream. We believed March 2020 had marked more than a temporary inflation shock. Something deeper had changed in the political economy that produced four decades of declining inflation and interest rates.

Governments had discovered how quickly the normal limits on intervention could disappear. Fiscal policy could put purchasing power directly into the economy. Central banks could stabilise markets on an extraordinary scale. Strategic competition could justify industrial policy. Security could outrank efficiency.

And all of that costs money.

Redundant factories are more expensive than concentrated ones. Domestic supply chains often cost more than imports. Strategic inventories consume capital. Defence, energy security and reindustrialisation require enormous quantities of labour, commodities and financing before producing anything useful.

As we wrote in August:

Resilience is valuable. It is rarely cheap.

That thinking shaped the Model Portfolio from the beginning. We were never betting that CPI would rise every month. We were positioning for a world in which inflation, interest rates and policy would travel across a wider and less predictable range, and where scarcity would periodically matter much more than investors raised during the 2010s had been conditioned to expect.

Two years later, very little about that argument sounds radical.

Fiscal dominance, defence spending, deglobalisation, energy security, industrial policy and structurally higher rates have become routine subjects of investment debate. Governments openly subsidise domestic manufacturing, semiconductors, power infrastructure and strategic resources. Markets now seem capable of finding an inflation angle in almost anything.

That last part is what interests us.

A geopolitical conflict becomes an energy-price story. Fiscal expansion becomes a Treasury-supply story. Strong growth becomes a reason for tighter monetary policy. Even artificial intelligence is often discussed first in terms of the electricity, semiconductors and capital required to build it.

None of those interpretations is absurd. Most contain a great deal of truth.

But when an explanation becomes the default explanation, investors should become uncomfortable.

Markets have a habit of taking yesterday’s difficult lesson and turning it into tomorrow’s lazy assumption.

And there is now evidence worth considering on the other side.

The first stage of the AI buildout is clearly resource-intensive. Data centres need processors, cooling, land, construction and extraordinary amounts of dependable power. It is competing for many of the same scarce inputs as defence, reindustrialisation and the rebuilding of energy infrastructure.

Yet we are spending all that capital to manufacture something whose price is collapsing:

Intelligence.

If machine intelligence allows businesses to write more software, perform more research, automate more processes and serve more customers without increasing labour and other inputs proportionately, productive capacity begins to expand in ways our original inflation framework must take seriously.

We do not know yet how large that effect will become.

That is precisely the point.

A thesis earns its keep by helping us interpret new evidence. It becomes dangerous when we start protecting it from that evidence.

August’s VMF’s Strategic Asset Allocation therefore does something deliberately uncomfortable. It revisits the Market View that has shaped Tier One since inception at the moment that view is receiving some of its strongest validation.

The excerpt below was first published for paid subscribers on 7 August 2026. It explains how the post-2020 regime emerged, why even some of the great defenders of the old deflationary order are changing their minds, and why that intellectual victory should make us more demanding rather than more complacent.

The investment problem has changed.

We no longer need to convince the market that the old regime is gone.

We need to find what the market might be missing now that almost everyone knows it.

When contrarian becomes consensus, the work starts again.

Good reading.

March 2020 did not necessarily begin the Fourth Turning.
It made the Fourth Turning investable.

Strauss and Howe describe a Fourth Turning as a crisis era in which the old institutional order is challenged, collective priorities displace individual preferences and society mobilises in response to threats it regards as existential. A First Turning follows the eventual resolution: an era of stronger institutions, greater social coordination and a new civic order. The framework is interpretive rather than mechanically predictive, but it offers a useful way to understand why economics, politics and markets can change together rather than independently.

The pandemic accelerated that transition because it revealed how quickly the normal limits on policy could disappear.

The Federal Reserve reduced its policy rate to the effective lower bound, expanded its securities purchases and deployed a broad collection of emergency facilities.

Fiscal authorities simultaneously provided transfers, income support, credit guarantees and business assistance. The scale and coordination of the intervention supported a rapid recovery, but it also placed powerful demand stimulus into an economy whose productive capacity remained impaired.

The immediate inflationary consequence was visible.
The more durable consequence was political.

Once governments demonstrate that they can mobilise fiscal and monetary resources on an extraordinary scale, the boundary of what future crises can demand changes. Policies previously regarded as exceptional enter the available toolkit.

Households, businesses and markets begin to incorporate a different reaction function.

The state becomes more willing to absorb risk.

The central bank becomes more closely entangled with financial stability and government financing.

The public becomes less willing to accept economic pain when large interventions appear possible.

That does not mean every future downturn will produce another identical stimulus programme.

It means the political constraint against intervention has weakened.

The pandemic also accelerated forces that had been developing before the first lockdown...

China and the United States were already moving from commercial integration towards strategic competition. Supply chains built around the lowest-cost producer were becoming politically uncomfortable. Energy systems designed primarily around price were being reconsidered through security and resilience. Defence expenditure was beginning to rise. Industrial policy was returning to economies that had spent decades trusting global markets to allocate production.

Covid transformed those tendencies into priorities.

Just-in-time became just-in-case.

Efficiency became redundancy.
Global supply became domestic capacity.
Commercial dependence became strategic vulnerability.

The economics of those choices are clear. Redundant factories cost more than concentrated ones. Domestic production frequently costs more than importing from the lowest-cost supplier. Strategic inventories tie up more capital than lean supply chains. Defence, energy security and grid resilience require large amounts of commodities, labour and financing before they generate useful output.

Resilience is valuable.
It is rarely cheap...

That was the central economic conclusion behind the Market View with which VMF Research began publishing.

The new regime was never simply about a larger money supply. It reflected a deeper change in the interaction between politics, policy and productive capacity. Governments can create financial claims. They cannot create engineers, power plants, transformers, semiconductor factories, strategic mines or skilled workers by decree. When fiscal and monetary demand expand faster than the economy’s ability to produce, prices become the rationing mechanism.

March 2020 brought those forces together.

The Fourth Turning may have begun with the institutional and financial crisis of 2008. That episode permanently expanded the role of the Federal Reserve’s balance sheet and established the central bank as the ultimate stabiliser of financial markets.

But much of the money created after 2008 remained inside the financial system.

Asset prices rose.
Consumer-price inflation did not.

The response to the Covid Crash was fundamentally different.

The Federal Reserve returned interest rates to zero and expanded its balance sheet at extraordinary speed. Fiscal authorities simultaneously transferred purchasing power directly to households and businesses. Demand was supported immediately, while productive capacity remained impaired.

Monetary and fiscal policy did not merely operate alongside one another.

They reinforced one another.

The charts throughout this section capture the significance of that moment. The federal funds rate collapsed as the budget deficit approached levels without modern peacetime precedent. The Federal Reserve’s balance sheet moved sharply higher. Ten- and thirty-year Treasury yields reached the lows from which they would begin reversing a four-decade secular trend.

No single market series can establish the beginning of a new regime.

Together, however, they identify an unmistakable break.

March 2020 marked the moment when policy support escaped the financial system and entered the real economy with sufficient force to collide with constrained supply.

That was why we believed inflation and interest rates would become higher and more volatile than they had been during the preceding decades. And that view shaped the Model Portfolio from its inception. The large Alternatives Book, the emphasis on precious metals and crypto assets, the exposure to natural resources inside the Equity Book and the deliberate underweight to conventional fixed income were all expressions of the same strategic conclusion.

We were not forecasting that inflation would rise every month. We were positioning for a world in which inflation, policy and interest rates would move across a wider and less predictable range (and in which the traditional relationship between stocks and bonds could become less dependable).

The old regime rewarded financial duration. The new one would reward scarcity, adaptability and real productive capacity.

Much of that view has since been validated.

Fiscal deficits remain large even outside recession. Globalisation is fragmenting. Governments are rebuilding defence, energy and industrial capacity. Supply-chain efficiency is increasingly being sacrificed for security. The electricity grid has become a strategic bottleneck. Capital requirements are rising across semiconductors, artificial intelligence, reindustrialization and national infrastructure.

Long-term interest rates have responded accordingly.

The forty-year decline in Treasury yields ended around the Covid Crash. Bond investors now demand greater compensation for inflation uncertainty, fiscal supply and term risk. Markets treat stronger growth with suspicion and interpret new government spending through its possible inflationary consequences.

What was once a contrarian Market View has become increasingly conventional...

That does not make it wrong.
But it changes the investment problem.

The greatest opportunities rarely appear when an important thesis is being ignored. They appear when a correct thesis becomes so widely accepted that every new development is forced into the same conclusion.

Our concern is therefore not that the post-2020 regime thesis has failed.

It is that it has become too easy to repeat.

The evidence of that migration is now difficult to ignore. Investors who spent decades warning about debt-driven deflation are reconsidering their frameworks. Analysts who once believed excessive leverage would keep inflation and long-term interest rates permanently subdued are beginning to argue that deglobalisation, capital scarcity and constrained productive capacity have changed the equilibrium.

The higher-and-more-volatile inflation regime is moving from the edge of macroeconomic debate towards its centre.

At the same time, the incoming evidence is becoming less one-sided.

Productivity is improving. Wage pressure appears less threatening when measured against output per worker. The cost of applying intelligence is falling rapidly. And the Federal Reserve itself is beginning to examine whether technology is expanding the economy’s productive capacity faster than its traditional frameworks can capture.

None of this invalidates the post-2020 regime.

A world of higher and more volatile inflation can still produce powerful periods of disinflation. Indeed, volatility implies movement in both directions. But it changes the question investors should be asking.

The original challenge was recognising that the disinflationary stability of the previous era had ended. The challenge today may be recognising that the market has learned that lesson so completely that it now interprets almost every geopolitical, fiscal and economic development through the same inflationary lens.

Our Market View has not failed.
It has moved towards consensus.

And when a formerly contrarian thesis becomes the market’s organising assumption, it must be tested more aggressively… not defended more emotionally.

The clearest evidence of how far this intellectual migration has travelled comes from one of the most respected deflationary thinkers of the past several decades.

Lacy Hunt spent much of his career explaining why excessive debt, falling velocity and weak demand would suppress inflation and long-term interest rates.

For a very long time, he was right.
Now, his view is changing.

To understand why the inflationary regime has become so widely accepted (and what the market may consequently be overlooking) we must first understand what changed his mind.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the August 2026 issue of VMF’s Strategic Asset Allocation. The research, views and Tier One Model Portfolio information are stated as of 7 August 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 7 August 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Tier One Model Portfolio is an illustrative research portfolio designed to communicate VMF Research’s strategic asset-allocation framework, Market View and investment conclusions. It does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines macroeconomic, monetary-policy, geopolitical, technological, market and strategic asset-allocation research within a medium- to long-term framework. Statements concerning inflation regimes, fiscal policy, interest rates, artificial intelligence, productivity, disinflation and potential changes in market leadership are analytical judgments and conditional hypotheses rather than assurances. Forecasts and scenario outcomes may change materially as economic data, policy, technology and market conditions evolve.

Investments discussed within VMF’s Strategic Asset Allocation may involve substantial risks, including market volatility, changes in inflation and interest rates, geopolitical disruption, regulatory intervention, currency movements, liquidity constraints, technological disruption and the possible loss of capital. Improvements in productivity or artificial-intelligence capabilities may not translate into lower inflation, stronger corporate earnings or superior investment returns. Likewise, inflationary, technical and valuation frameworks may fail.

Disclosure of interests: as of the research cut-off, legal entities controlled by Vasco Marques de Freitas held long positions in the KraneShares CSI China Internet ETF (KWEB), ARK Innovation ETF (ARKK) and Malibu Life Holdings Limited, as well as listed exchange-traded products providing economic exposure to Bitcoin, Ether and Solana. These interests may create actual, potential or perceived conflicts and should be considered when evaluating the analysis, recommendations and Model Portfolio conclusions. The existence of an interest does not validate a recommendation, diminish its risks or imply suitability for any particular reader.

Neither VMF Research nor the author received compensation from any issuer, fund sponsor or other entity covered in connection with the preparation of this research, and no covered entity reviewed, approved or amended its investment conclusions before first dissemination.

Past performance, Model Portfolio performance and forward-looking scenarios are not reliable indicators of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

AI has a dirty little secret...

The smarter the digital world becomes, the more brutally physical its requirements become.

Models can improve in weeks. Software can be copied in seconds. Inference costs can collapse almost overnight.

Power plants cannot.

Transmission lines cannot.

And uranium mines certainly cannot.

That contradiction may become one of the defining investment problems of the next decade.

The market has spent enormous amounts of time debating which model will win, which chip will dominate and which platform will capture the next wave of AI spending.

But every one of those arguments eventually runs into the same wall:

Electricity.

A data centre does not care how elegant the software is if the grid cannot keep it running.

And for workloads that must operate around the clock, availability is not a secondary feature. It is the product.

That is where nuclear power enters the story.

Not because AI suddenly created the uranium bull case.

It did not.

The uranium market was already tightening. The existing global reactor fleet consumes more uranium than primary mines produce each year. Reactor life extensions, restarts and new construction were already pushing future requirements higher. Utilities had already spent years contracting less material than they would ultimately need.

AI simply makes the bottleneck harder to ignore.

It adds a new source of electricity demand to a fuel market that was already living on borrowed inventory.

And something more interesting is beginning to happen.

Governments and state-linked buyers are moving before the demand arrives.

India is securing uranium for a nuclear fleet it has not yet built. China has become a major force in long-term contracting. Kazakhstan, despite being the world’s largest producer, has moved towards building strategic inventories of its own.

The buyers are moving before the miners.

That changes the character of the commodity.

When countries begin accepting the carrying cost of securing future supply because the consequences of not having it are far greater, uranium stops behaving like an ordinary industrial input.

It starts behaving like a strategic reserve asset.

That matters because the supply response is painfully slow.

The uranium price can move in an afternoon.

A mine may take a decade.

Permitting, geology, financing, processing infrastructure and politics all stand between a higher commodity price and an additional pound of production. Even established mines routinely disappoint.

Meanwhile, reactors do not casually switch themselves off because fuel became more expensive.

Uranium represents only a fraction of the economics of operating a nuclear plant. Once the reactor exists, security of supply matters far more than saving a few dollars on the fuel bill.

That asymmetry is what makes the setup interesting.

Supply is slow.

Demand is stubborn.

Inventories are finite.

And the world is asking nuclear power to do more.

July’s Alpha Tier followed our AI thesis out of the digital economy and into the physical constraint underneath it. The same abundance of intelligence that may pressure one scarcity can intensify another.

More useful intelligence can mean more automation.
More automation can mean more computation.
More computation means more electricity.
And electricity needs fuel.

The excerpt below was first published for paid Alpha Tier subscribers on 24 July 2026 at 9:00 p.m. Eastern Daylight Time. It examines why uranium is becoming strategically more important, why years of deferred procurement matter, and why a market capable of repricing in weeks is dependent on supply that may require years to respond.

The AI boom may ultimately be remembered for algorithms.

But algorithms do not keep the lights on.

The fuel beneath the boom does.

Good reading.

Reliable power has a fuel.

Nuclear generation operates continuously, at scale and with far less dependence on weather than most competing sources of electricity. US energy data place its capacity factor at approximately 91% (materially above natural gas, coal, hydro, wind or solar). For data centres running valuable workloads around the clock, availability is not an attractive feature.

It is the product.

The demand side is becoming more visible. Global data-centre electricity consumption could rise approximately two-and-a-half times by 2030, while AI related power use may grow even faster. Renewables will supply part of that increase, but intermittency, grid congestion and the slow expansion of storage leave an essential role for dependable generation. Nuclear power is increasingly being treated as one of the few scalable answers.

Yet the uranium case does not require every AI electricity forecast to prove correct.

The existing reactor fleet already consumes more uranium than the world mines each year. Life extensions, restarts and new construction were strengthening demand before the latest data-centre boom. Artificial intelligence adds urgency, political support and a new class of well-capitalised power buyers.

It accelerates a shortage that was already forming.

The clearest recent evidence comes from India. The country currently operates approximately 8 gigawatts of nuclear capacity but is targeting 100 gigawatts by 2047. Reaching that objective will require sustained construction over two decades... and considerably more fuel than India can produce domestically.

Its procurement strategy is moving ahead of the reactors. A recently agreed contract with Canada’s Cameco covers almost 22 million pounds of uranium from 2027 through 2035. India and Australia have also completed the administrative arrangements required to permit Australian uranium exports under their existing nuclear-cooperation framework. Australia controls roughly 28% of known global uranium reserves, opening another potentially important supply route to one of the world’s largest future consumers.

So, India is securing fuel for generating capacity that has not yet been built.
It is not alone.

China represented roughly half of global long-term uranium contracting in 2024. Kazakhstan, already the world’s largest producer, has moved towards building strategic inventories of its own. Sovereign and state-linked buyers are increasingly competing with utilities for material whose future availability can no longer be assumed.

The buyers are moving before the miners.

The important signal is the order of events. Fuel is being secured before the corresponding demand fully reaches the market. Buyers are accepting the carrying cost of early procurement because the potential cost of insufficient supply is far greater than the benefit of waiting for a slightly lower price.

Uranium is beginning to behave less like an ordinary industrial input and more like a strategic reserve asset.

The market is poorly positioned for that transition.

World mine production already falls short of the annual requirements of the existing reactor fleet. For decades, the difference was supplied through inventories, military stockpiles, reprocessing, underfeeding and other secondary sources created partly by the enormous overproduction of the Cold War.

Those buffers postponed the adjustment. They did not create new mines.

The world now has 440 operational reactors, with 79 under construction and another 120 planned. The World Nuclear Associations' reference scenario projects annual reactor requirements rising from approximately 175 million pounds of U₃O₈ equivalent in 2024 to 391 million pounds by 2040.

Not every proposed reactor will be completed. Timetables will slip, policies will change and construction costs will remain difficult to control. The broad direction is nevertheless clear. Demand is expected to rise from a market that cannot fully supply today’s consumption from primary production.

Supply-and-demand scenarios compiled from UxC and Cameco data indicate a cumulative uranium shortfall of approximately 1.4 billion pounds through 2045 under the base case. If the international objective of tripling global nuclear capacity by 2050 were achieved, the projected deficit would exceed 3 billion pounds.

These forecasts are not promises. Their value lies in illustrating the scale of the response required. Balancing the market would demand a combination of substantially higher mine output, additional secondary supply, lower reactor growth or sustained inventory depletion. None is assured.

The textbook answer to a commodity shortage is higher prices. Higher prices encourage production, attract capital and eventually restore equilibrium.

Uranium complicates that sequence.

The commodity price can move in weeks. A mine may require a decade. Resources must be discovered, delineated and permitted. Processing infrastructure must be financed and constructed. Governments can alter royalties, taxes or operating conditions. Technical studies must survive contact with geology, and projects that appear economic on paper can suffer delays, dilution and cost overruns before producing a single pound.

Even established operations regularly miss production targets.
Supply therefore responds slowly and unpredictably.

Demand is less flexible.

Uranium represents a relatively small share of the total cost of operating a nuclear power plant. Once the facility exists, its owner has little economic incentive to interrupt generation in order to save a few dollars on fuel. The reliability of supply is considerably more important than obtaining every pound at the lowest available price.

Utilities have nevertheless spent much of the past decade postponing the decision. Long-term contracting has remained below replacement requirements for 13 consecutive years. Approximately 116 million pounds were contracted in both 2024 and 2025, materially below annual reactor demand. Even the apparent return to replacement-level contracting in 2023 was distorted by a large one-off purchase from Ukraine.

The market has deferred the purchases.
It has not removed the reactors.

As older contracts expire and available inventories decline, utilities must eventually return to the term market (or accept greater exposure to a comparatively thin spot market that may not contain the volumes they require).

The geographical structure of supply makes waiting more dangerous. Kazakhstan accounted for approximately 39% of world uranium production in 2024. Canada represented another 24%, while Namibia, Uzbekistan and Australia provided much of the remainder. A portion of Kazakhstan’s output is transported through Russia, which also remains central to conversion and enrichment (the specialised stages required to transform mined uranium into usable reactor fuel).

Ore in the ground is not the same as assured fuel in a reactor. Mining, conversion, enrichment, fabrication and transport form one interconnected supply chain. A bottleneck at any point can tighten the entire system, raise inventory requirements and increase the strategic value of supply from politically dependable jurisdictions.

The uranium cycle is therefore being shaped by two forces at once:

• More reactors require more fuel.

• Deglobalisation requires more secure fuel.

Western governments are attempting to rebuild capabilities that were allowed to erode during decades of inexpensive imports. The United States, once the world’s dominant uranium enricher, now has no domestically owned commercial enrichment capacity and has lost almost all of its mine production. Reconstructing that chain will require capital, time and prices capable of rewarding the investment.

The path from a tighter uranium market to uranium equities is not automatic.

It is, however, powerful when it works.

Higher long-term prices expand the margins of existing producers. Better cash flow supports mine extensions, restarts and brownfield development. Resources that were uneconomic at lower prices become more valuable, improving financing conditions for developers and raising the strategic value of credible projects.

The operating leverage can be substantial.

Capturing it without allowing one mine, jurisdiction or management team to determine the outcome is the more difficult part.

Physical uranium offers the cleanest exposure to the commodity, but none of the earnings leverage available to producers. A single miner can deliver far greater upside, yet the result may depend on one asset performing exactly as expected. Junior miners increase the optionality further, together with the financing, dilution and execution risks attached to projects that may remain years from production.

URNM ( ) offers the more balanced expression.

The Sprott Uranium Miners ETF (URNM) combines established producers, developers and physical uranium within one vehicle. As of 30 June, it held 26 issuers, with approximately 82% invested in uranium-related equities and 18% in physical uranium. The structure therefore captures higher fuel prices, expanding producer margins and the increasing strategic value of credible future supply without making the thesis dependent on one mine succeeding.

Tier One subscribers will already know the position. URNM has been held in VMF’s Strategic Asset Allocation Model Portfolio since August 2024. The original recommendation performed strongly before the ETF surrendered a substantial part of its advance over the past several months.

That correction has not weakened the uranium thesis.
But it has restored the asymmetry.

Since inception, URNM has maintained a constructive relative-strength trend against the broader US equity market. That long-term market confirmation lends credibility to the fundamental case, even though the path has been exceptionally volatile. The short-term picture is now far less comfortable. Since the beginning of the year, the ETF has fallen almost 45%, broken below its shorter-term trend measures and returned to an important support zone.

Momentum is weak. But the long-term structure remains intact. That combination is precisely why the position now merits consideration inside Alpha Tier.

There is an obvious parallel with our China exposure. In both cases, the fundamentals are improving faster than the price action. But the portfolio implications are different. China already represents a meaningful overweight, and adding more would deepen an existing concentration before the broad technical evidence confirms the thesis.

You might also like reading:

URNM diversifies the portfolio into another sector, another set of earnings drivers and another physical bottleneck. Its short-term technicals are equally damaged, but its long-term relative-strength structure remains considerably more constructive.

We are therefore willing to act before a full technical recovery.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the July 2026 issue of Alpha Tier. The research, views and Alpha Tier Model Portfolio information are stated as of 24 July 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 24 July 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Alpha Tier Model Portfolio is an illustrative research portfolio and does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines thematic, fundamental, commodity, energy-market, geopolitical, technical and portfolio-construction research within a medium- to long-term investment framework. Statements concerning future electricity demand, nuclear-power deployment, uranium requirements, contracting activity, mine supply and the potential effects of artificial intelligence on energy consumption are analytical judgments and scenario-based estimates rather than assurances. Forecasts may change materially as reactor construction, energy policy, commodity markets and supply conditions evolve.

Investments in uranium, uranium-mining companies and related exchange-traded funds involve substantial risks, including commodity-price volatility, mine-development and operating risk, permitting delays, cost inflation, financing and dilution risk, political and jurisdictional uncertainty, changes in nuclear-energy policy, reactor delays or cancellations, supply responses, inventory releases, geopolitical disruption, currency movements, liquidity constraints and the possible loss of capital. Higher uranium prices or greater nuclear-power demand may not translate into higher earnings or investment returns for uranium-related companies or funds. Technical and relative-strength signals may also fail.

Neither VMF Research nor the author received compensation from any issuer or other entity discussed in connection with the preparation of this research, and no issuer reviewed, approved or amended its conclusions before first dissemination.

Past performance, Model Portfolio performance and commodity-price history are not indicative of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

The strongest thesis in the room can still be the wrong place for the next dollar.

That is one of the least intuitive lessons in investing.

New evidence arrives. Confidence rises. The temptation is obvious: buy more.

But portfolios are not conviction contests.

In The Boom Changes Owners, we argued that China’s rapidly improving AI ecosystem is making our broader thesis harder to dismiss. Capable intelligence is becoming cheaper. Competition is widening. And as models become more interchangeable, value should increasingly migrate towards the businesses controlling customers, proprietary data, distribution and commercial workflows.

The evidence has strengthened.

So has our conviction.

Yet Alpha Tier is not increasing its China weight.

Why?

Because a good portfolio decision cannot be made by asking whether a thesis looks better than it did yesterday. The relevant question is whether the next unit of capital offers better asymmetry there than anywhere else.

That distinction matters enormously.

Existing exposures can overlap. Diversified vehicles can conceal substantial company-level concentration. And a position that was attractive at one weight can become much less attractive when the portfolio already carries enough of the underlying economic bet.

There is another constraint too.

Price.

China’s technological progress is moving faster than the market’s willingness to reward it. Some of the fundamental evidence is increasingly persuasive. The broader technical confirmation remains incomplete.

That leaves us in an unusual position:

We are more confident in the thesis than we were before.
We are not yet more confident in adding capital to it.

This is where investment research becomes portfolio management.

Finding a good idea is only the beginning. The harder discipline is deciding how much of it deserves to be owned, what evidence should justify adding more, and when the next dollar belongs somewhere else.

Or, as we wrote in July’s Alpha Tier:

Fundamentals tell us what deserves to be owned.
Price tells us when conviction deserves more capital.

The excerpt below was first published for paid Alpha Tier Subscribers on 24 July 2026 at 9:00 p.m. Eastern Daylight Time. It explains why stronger evidence behind our China thesis led us to maintain, rather than increase, the existing allocation.

Sometimes conviction tells you to buy.

Sometimes it tells you to wait.

The difference is what separates an idea from a portfolio.

Good reading.

The obvious response to stronger evidence would be to buy more China.

We are not going to do that.

The Alpha Tier Model Portfolio already owns MCHI, KWEB, Alibaba and PDD. That headline understates the degree of concentration. Alibaba is also one of the two largest constituents in both MCHI and KWEB, meaning our economic exposure to the company is materially greater than the direct BABA allocation alone suggests.

That concentration is deliberate.

Alibaba is one of the few businesses capable of participating across almost the entire Chinese AI value chain. It develops the Qwen model family, operates one of the country’s leading cloud platforms and controls the commercial ecosystems through which cheaper intelligence can be deployed at scale. Its commerce, advertising, logistics and enterprise businesses provide customers, proprietary data and economically valuable workflows that independent model laboratories must still find elsewhere.

The latest evidence therefore strengthens an important conclusion.

Alibaba does not need Qwen to win every benchmark.

It needs the supply of capable intelligence to expand, the cost of deploying it to decline and Chinese enterprises to integrate it into a wider range of commercial activity. Alibaba can benefit by developing its own models, distributing third-party models, selling the computing capacity required to run them and applying AI throughout businesses that already serve hundreds of millions of users.

Few companies occupy so many favourable positions at once.

That is why our true exposure to Alibaba matters more than the direct position shown in the Model Portfolio table.

The look-through analysis will make two things clear.

First, Alibaba is already one of the Model Portfolio’s more consequential individual company exposures.

Second, any decision to increase MCHI or KWEB would also increase our effective Alibaba position, even if the trade were presented as an addition to diversified Chinese equities.

We must therefore distinguish between two separate questions.

Has the latest evidence increased our confidence in Alibaba and the wider Chinese application-layer thesis?

Yes.

Has it improved the asymmetry enough to justify greater concentration before the market confirms that interpretation?

Not yet.

The fundamental case nevertheless deserves to be understood because Alibaba’s position inside the Chinese AI ecosystem is unusually difficult to replicate. Independent laboratories may produce exceptional models. DeepSeek has established extraordinary cost efficiency. Kimi and GLM have pushed closer to the frontier in coding and agentic tasks. MiniMax has developed competitive multimodal capabilities. Tencent owns enormous distribution through Weixin and a substantial cloud and gaming ecosystem.

But most of these businesses occupy only part of the chain.

Alibaba is closer to a full-stack expression.

It owns models, cloud infrastructure, enterprise relationships, consumer distribution, transactional data and the applications capable of converting better intelligence into commercial output. It can participate when Chinese companies train more models, when they consume more inference, when enterprises move workloads into the cloud and when AI improves search, advertising, merchant productivity, customer service and logistics inside Alibaba’s own ecosystem.

The distinction is especially important in a market where model leadership may prove temporary.

A laboratory can lose its benchmark advantage with the next release. A model that appears differentiated today can face pricing pressure within months.

The competitive position of a platform controlling customers, data and distribution is more durable.

Alibaba does not need to forecast the permanent winner at the model layer. It can help distribute the winner.

This does not mean Alibaba dominates every category.

DeepSeek may retain superior cost economics. GLM and Kimi may outperform Qwen on individual workloads. Tencent may possess stronger distribution in social media. Independent laboratories may innovate faster because they are less constrained by the priorities of a large organisation.

The investment case does not require otherwise.

It rests on Alibaba’s ability to monetise an expanding ecosystem rather than own every technical breakthrough within it. As intelligence becomes cheaper and model choice proliferates, the capacity to route workloads, serve enterprises and embed AI into established commercial systems may prove more valuable than temporarily controlling the highest score on one leaderboard.

The infrastructure data reinforce that opportunity.

China’s leading cloud platforms are still investing at a fraction of the scale of their American counterparts. Goldman Sachs estimates that capital expenditure by major US cloud providers rose from approximately $254 billion in 2024 to $443 billion in 2025 and could exceed $1 trillion in 2027. Equivalent spending by the principal Chinese platforms increased from approximately $36 billion to $57 billion and is projected to reach $123 billion over the same horizon.

The difference is enormous.

But it should not be interpreted simplistically.

China’s smaller capital programme partly reflects a less developed cloud market and constrained access to the most advanced computing hardware. Spending less is not automatically evidence of superior efficiency, particularly if the smaller investment base limits training capacity or slows the deployment of new infrastructure.

Yet the comparison also reveals how much runway remains.

Chinese model capability has advanced rapidly before domestic AI investment has approached anything resembling Western scale. If improving models stimulate cloud adoption, enterprise deployment and agentic workloads, Alibaba and its peers retain considerable room to expand their infrastructure.

The financing comparison is as important as the absolute spending.

Capital expenditure by the principal US hyperscalers is estimated to approach or exceed their combined operating cash flow during the most intensive years of the buildout. The equivalent ratio for Chinese platforms is also rising, but from a much lower starting point and towards a less extreme peak.

That difference does not eliminate risk. Chinese companies can misallocate capital, invest too aggressively or fail to convert infrastructure into attractive returns. But it suggests that China’s leading platforms can increase AI investment without placing the same immediate pressure on cash generation as several US hyperscalers.

China therefore appears to possess both runway and financial capacity.

The harder question is whether Alibaba can transform that investment into revenue.

That is the real execution test.

Capital expenditure alone does not create shareholder value. It creates the capacity from which value may eventually emerge. Alibaba must demonstrate that additional infrastructure attracts enterprise workloads, supports greater model consumption and produces enough incremental cloud revenue to justify the investment.

The forecasts are encouraging.

Goldman Sachs expects Alibaba’s rolling incremental cloud revenue relative to lagged capital expenditure to improve steadily through 2027. By the end of the forecast period, the company’s conversion ratio could exceed the comparable estimates for Google and Amazon.

That would represent a material change. Alibaba’s cloud business has historically possessed the infrastructure, customer relationships and strategic relevance without consistently delivering the operating leverage investors expected. A sustained improvement in capital-to-revenue conversion would indicate that the AI cycle is beginning to translate technical progress into financial output.

The wider capital-expenditure outlook suggests that this transition may extend beyond Alibaba.

Chinese internet investment is expected to continue rising through 2028, but at a progressively slower rate after the initial surge.

That is not a weakness in the thesis.
It is what the next phase should look like.

The industry is moving from building capacity towards proving what that capacity can earn.

Alibaba is the Model Portfolio’s most integrated expression of that transition. It develops models, owns cloud infrastructure and controls the commercial ecosystems through which cheaper intelligence can be converted into revenue. But it is not our only exposure to the application layer.

PDD Holdings offers a purer downstream expression. It does not need to train the leading Chinese foundation model or finance a hyperscale computing platform. It needs access to increasingly capable intelligence at a declining cost and the ability to deploy it across assets that remain scarce.

Those assets already exist. PDD controls merchant relationships, transaction data, consumer distribution and an expanding international network. Cheaper intelligence can improve product discovery, advertising conversion, fraud detection, translation, customer service, inventory management and logistics without requiring PDD to own the model producing every improvement.

The model may become commoditised. PDD’s network does not.

You might also like reading

KWEB provides the broader internet exposure. Its constituents span commerce, cloud computing, digital advertising, software, entertainment and online services, allowing the portfolio to participate if the benefits of cheaper intelligence diffuse across the Chinese platform economy. That breadth is useful, but it is not as diversifying as the ETF label initially suggests. KWEB contains several companies we also own directly (most importantly Alibaba).

MCHI serves a different purpose. It captures the possibility that technological progress contributes to a wider reassessment of Chinese assets through stronger investment, productivity and confidence. Yet it, too, adds to our effective Alibaba exposure.

The roles are therefore distinct. Alibaba is the integrated model, cloud and commercial-platform position. PDD is the downstream application. KWEB captures the wider internet ecosystem. MCHI expresses the possibility that technological progress contributes to a broader Chinese-equity recovery.

The latest evidence increases the expected value of all four.

It does not automatically increase their appropriate Model Portfolio weights.

That distinction is especially important because our direct positions understate the concentration already embedded in the portfolio. Alibaba is also one of the largest holdings inside MCHI and KWEB. Additional capital allocated to either ETF would therefore reinforce an existing company-level exposure rather than create an entirely new source of return.

A stronger thesis can make what we already own more valuable without making further concentration desirable. The next unit of capital must compete against every opportunity available to the Model Portfolio... not merely against an increasingly attractive China narrative.

Fundamentals tell us what deserves to be owned.
Price tells us when conviction deserves more capital.

The charts are not there yet.

The MCHI chart on the previous page shows the scale of the damage that still needs to be repaired. Chinese equities have underperformed the broader US market dramatically for more than a decade, and the ETF remains only modestly above the levels reached in 2015.

Few charts provide a clearer record of the collapse in investor confidence. The decline from the 2021 peak to the 2022 low was particularly consequential. It broke the previous rising structure and reset the market onto a far lower trajectory. That collapse absorbed a remarkable concentration of negative forces: the consequences of prolonged Covid lockdowns, real-estate deflation, balance-sheet stress, regulatory intervention and the widening geopolitical conflict between China and the United States.

Investors have not lacked reasons to avoid the market. MCHI is the cumulative record of them. A new upward channel may now be developing from the 2022 low, but the recovery remains incomplete. The ETF has retreated towards the lower half of that structure and remains below its 50-week moving average.

The market has begun to stabilise. But it has not yet demonstrated that improving technology fundamentals can overcome the wider macroeconomic and geopolitical discount.

KWEB tells the same story in a more concentrated form.

Its collapse after 2021 represented far more than the unwinding of an expensive valuation. It reflected a wholesale reassessment of regulation, profitability, capital allocation, geopolitical risk and the investability of China’s digital economy.

The recovery remains uneven. KWEB is still far below its previous highs, beneath a declining 50-week moving average and close to its 200-week moving average. Its relative performance against the broader US equity market continues to weaken.

That is what makes the present divergence so compelling. Chinese model capability is improving. Intelligence costs are falling. Cloud platforms retain substantial room to invest. Yet the market continues to price much of the internet complex as though those developments will produce little durable value.

The mismatch is unusually large.
But valuation is not timing.

Markets can resist better fundamentals for far longer than expected when regulation, geopolitics and confidence remain unresolved. KWEB still requires evidence that buyers are prepared to move beyond isolated opportunities and reengage with the wider sector.

Alibaba is beginning to provide some of that evidence.

The stock has outperformed KWEB since 2024, helping validate our decision to own Alibaba directly rather than rely exclusively on the diversified ETF.

That relative strength is consistent with the analysis developed throughout this issue. Investors are beginning to distinguish Alibaba’s combination of models, cloud infrastructure, financial resources, distribution and commercially valuable workflows from the broader Chinese internet complex.

The chart is more constructive than KWEB’s, but it is not fully resolved.

Alibaba remains inside its rising channel and above its 200-week moving average. The recent correction, however, has taken the stock below its shorter-term trend measures and interrupted the previous momentum.

The market is rewarding Alibaba’s superior positioning.

But it is not yet doing so decisively enough to justify increasing an exposure that is already substantial on a look-through basis.

PDD is approaching a different kind of decision.

Its valuation may be even less demanding, but the chart will attract few momentum investors.

PDD remains below its 20- and 50-week moving averages and beneath the descending trendline extending from the all-time high. At the same time, it is approaching long-term support from the rising trendline drawn from the 2022 low.

Those two forces are converging. The result is a compressed technical structure between persistent overhead resistance and a rising long-term floor. An upside resolution would suggest that the market is beginning to recognise PDD’s ability to translate cheaper intelligence into better search, advertising, merchant services, localisation and logistics. A break of support would indicate that the fundamental case still lacks sufficient sponsorship.

We believe the setup can ultimately resolve higher as the wider Chinese equity complex is reassessed.

But the chart is approaching confirmation.
It has not delivered it.

That is the conclusion across the Model Portfolio’s four China positions. The fundamental evidence has strengthened, but price confirmation remains narrow and inconsistent. Alibaba is the strongest expression. MCHI is attempting to establish a new recovery. KWEB still reflects profound scepticism. PDD remains inside a consequential decision zone.

That is sufficient reason to maintain the overweight.
It is not sufficient reason to increase it.

We will therefore keep MCHI, KWEB, Alibaba and PDD at their current weights.

Additional capital should require evidence that the improvement is broadening beyond one company and becoming visible across the market itself.

At a minimum, we want renewed strength from either MCHI or KWEB, supported by confirmation from at least one direct holding. Until then, the Model Portfolio already owns enough of the upside... and enough of the risk.

The analysis does, however, point towards another opportunity.

If cheaper and more capable models lower the cost of useful automation, the number of tasks attempted across the economy should expand. Each individual task may require less computation, but the growth in models, agents and automated workflows could more than offset those efficiency gains.

More intelligence means more computation.
More computation means more data centres, networks and cooling.
Above all, it means more electricity.

We are therefore not allocating the next unit of capital to an existing China overweight before the market confirms the thesis. We are following the same economic transition into the physical system required to sustain it... towards a constraint that better software cannot remove and supply cannot expand quickly.

Intelligence is becoming more abundant.
Reliable power is not.

You might also like reading:

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the July 2026 issue of Alpha Tier. The research, views and Alpha Tier Model Portfolio information are stated as of 24 July 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 24 July 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Alpha Tier Model Portfolio is an illustrative research portfolio and does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines thematic, fundamental, technological, valuation, technical and portfolio-construction research within a medium- to long-term investment framework. Statements concerning Chinese artificial intelligence, declining intelligence costs, cloud investment, application-layer economics and future monetisation are analytical judgments rather than assurances. Model Portfolio weights also reflect concentration, overlapping exposures, diversification, price behaviour and relative opportunity. A strengthening investment thesis does not necessarily imply that a larger allocation is appropriate.

Investments in Chinese equities, emerging markets, internet and technology companies and China-focused exchange-traded funds involve substantial risks, including government intervention, regulatory change, geopolitical tensions, restrictions on technology or capital flows, governance concerns, currency movements, intense competition, technological disruption, valuation compression, liquidity constraints and the possible loss of capital. Improvements in AI capability or adoption may not translate into higher revenues, earnings or investment returns. Technical and valuation signals may also fail.

Disclosure of interests: as of the research cut-off, neither VMF Research, Lda. nor Vasco Marques de Freitas personally held any financial instrument included in the Alpha Tier Model Portfolio. A legal entity controlled by Vasco Marques de Freitas held a long position in the KraneShares CSI China Internet ETF (KWEB), which is discussed in this article. This interest may create an actual, potential or perceived conflict of interest and should be considered when evaluating the analysis. The existence of an interest does not validate a recommendation, diminish its risks or imply that it is suitable for any reader.

Neither VMF Research nor the author received compensation from any issuer covered in connection with the preparation of this research, and no issuer reviewed, approved or amended its conclusions before first dissemination.

Past performance is not indicative of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

The market has mistaken an arms race for a monopoly.

For much of the past two years, artificial intelligence has been priced as though control of capable intelligence would remain concentrated inside a small group of Western laboratories and the companies supplying their infrastructure.

Capital followed that assumption.

Processors became strategic assets. Memory became a bottleneck. Data centres became factories for a new industrial revolution. The businesses receiving the cheques enjoyed immediate revenues, expanding margins and the comfort of visible demand.

China is now making that certainty dangerous.

DeepSeek opened the door. GLM-5.2 showed that the capability gap was narrowing. Kimi K3 has delivered a more unsettling message: Chinese models may be moving beyond cost leadership and beginning to compete for the economics attached to more valuable work.

Kimi is not the cheapest Chinese model.
That is what makes it important.

Customers appear willing to pay more when a system writes better code, completes longer workflows or produces a reliable result with fewer attempts. Yet Kimi remains less expensive than the leading Western frontier alternatives.

This is no longer a story about cheap imitation.
It is a market-structure story.

China is beginning to compete across both ends of the intelligence market: driving down the price of capabilities that are becoming common, while moving closer to the frontier where newly unlocked capabilities can still command a premium.

And it is no longer dependent on one exceptional laboratory.

DeepSeek, Kimi, GLM, Qwen and MiniMax now represent different organisations, architectures and commercial strategies. Individual leaders will change. Benchmarks will be overtaken. Some technically impressive model developers may still become terrible investments.

The broader signal is harder to dismiss.

China is building a competitive ecosystem capable of making useful intelligence better, cheaper and more widely available.

That gives new weight to a thesis we have been developing across VMF’s Market View.

In The Abundance Shock, we argued that falling intelligence costs would not eliminate scarcity. They would move it. The China AI Reset then examined the force accelerating that transition. The Intelligence Dividend followed cheaper intelligence into healthcare, where proprietary biological evidence remains difficult to reproduce. The Discovery Machine and The Royalty Engine showed how the same framework can ultimately reach individual businesses.

A common thread runs through all of them:

When intelligence becomes easier to obtain, the assets around it become more important.

Customers.
Distribution.
Proprietary data.
Embedded workflows.
Trusted relationships.
Physical infrastructure.

Those assets cannot be downloaded alongside a new model.

That is why the China technology thesis is gaining traction where a serious thesis should gain traction first: inside the mechanism.

The leading Chinese platforms do not need permanent control of the smartest model. They need a growing supply of capable intelligence, falling deployment costs and more businesses willing to integrate AI into commercial activity.

A platform controlling merchants, consumers, cloud infrastructure, transaction data and established workflows can benefit from a model it develops itself, a model supplied by a competitor or a model whose present benchmark advantage disappears six months later.

The model may become interchangeable.
The commercial ecosystem does not.

The market is still watching the leaderboards. We are watching who owns the customer when intelligence becomes cheap.

That distinction could reshape the AI profit pool. Infrastructure demand may remain enormous, particularly if lower costs generate an explosion in usage. But strong demand does not guarantee that every supplier preserves its scarcity premium, margins or valuation.

The companies financing the infrastructure may eventually capture more of the value it produces. Beyond them sit businesses capable of turning cheaper intelligence into better products, lower costs and deeper customer relationships.

That is how a boom changes owners.

The excerpt below was first published for paid Alpha Tier subscribers on 24 July 2026. It examines the latest Chinese model releases, the widening competitive ecosystem and why the application-layer thesis is becoming more credible even before the broader Chinese equity market has fully confirmed it.

The full issue then confronts the harder Alpha Tier question.

What should the next unit of capital own when the thesis is strengthening, but the portfolio already carries meaningful exposure to it?

This article begins with the evidence.
The boom is still growing.
The ownership of its profits may be changing.

Good reading.

The latest evidence arrived almost before the ink was dry.

When July’s VMF’s Strategic Asset Allocation was completed, GLM-5.2 provided some of the clearest evidence yet that China could perform a growing share of economically valuable work without reproducing the cost structure of the leading Western systems.

Days later, Kimi K3 moved the argument forward again.

The new open-weight model appears to compete near the global frontier in coding and selected agentic tasks. More importantly, it is priced materially above several leading Chinese alternatives while remaining cheaper than the principal Western frontier systems.

That combination matters more than another temporary position on a leaderboard.

DeepSeek showed that China could make intelligence cheaper. GLM-5.2 showed that the capability gap was narrowing. Kimi K3 may now be showing that the strongest Chinese models can command higher prices when they perform more valuable work.

China is therefore no longer competing only by offering yesterday’s intelligence at a discount. It is beginning to compete for the frontier itself... and for the economics attached to it.

Kimi is not the cheapest model in the comparison.

That is precisely the point.

Once a capability becomes common, its price should fall. Routine summarisation, translation, content generation and basic coding assistance will become increasingly difficult to differentiate. Providers competing only on capabilities that are already widely available will eventually be forced to compete on cost.

But intelligence is not one uniform product.

A model capable of completing more difficult work, producing more reliable code or navigating a longer agentic workflow can still command a premium. Customers do not ultimately purchase tokens.

They purchase useful results.

A cheap model that fails repeatedly can prove expensive. A more capable model that completes an economically valuable task on the first attempt may justify a higher advertised price while still producing a lower total cost.

The likely outcome is therefore not the complete commoditisation of every model. It is a more competitive market in which established capabilities become cheaper, while temporary pricing power migrates towards newly unlocked and commercially valuable work.

China is beginning to compete at both ends of that market.

The broader intelligence rankings reinforce the conclusion.

Kimi is not sitting alone at the edge of an otherwise Western frontier. GLM, Qwen, DeepSeek, MiniMax and other Chinese model families increasingly occupy positions throughout the leading group.

One strong release can be dismissed as an anomaly. A growing cohort is more difficult to ignore.

The significance of the chart above is not simply that China now has more models on global leaderboards. It is that the improvement extends across general text, coding and agentic workloads.

That breadth changes the probability distribution.

Individual laboratories will still rise and fall. Leadership can disappear with the next release, and competition may make many model developers poor investments even as their technology improves. Lower-end API pricing remains under pressure, training costs continue to rise and the capital required to remain near the frontier is substantial.

But the application-layer thesis does not require one permanent winner at the model layer. It requires competition to make capable intelligence better, cheaper and more widely available.

Deployment can reinforce that process. The research underlying this issue suggests that AI-generated code may already account for a very high proportion of output inside selected Chinese mega-cap technology companies. Those figures should not be treated as representative of the entire Chinese economy, but they illustrate the potential feedback loop: usage produces real-world successes, failures and corrections → that information improves post-training → better models then encourage more usage.

China may therefore be moving beyond the phase in which its laboratories learn principally from a frontier established elsewhere.

Increasingly, they can learn from their own users.

Its emerging domestic technology stack adds another layer of resilience.

Meituan’s LongCat 2.0 was reportedly trained and deployed on a large domestic computing cluster. This does not establish that Chinese accelerators have surpassed the leading American hardware. The more relevant point is that China appears increasingly capable of combining domestic silicon, sparse architectures, software optimisation and enormous engineering scale into a viable end-to-end system.

Export restrictions remain a constraint. But they are also influencing the form of Chinese progress, forcing laboratories to optimise models, hardware and deployment as an integrated system rather than rely indefinitely on access to one irreplaceable foreign component.

Constraints have not stopped progress.

They have shaped it.

That provides further evidence for a working hypothesis we have been developing for several months.

In The Abundance Shock, we argued that artificial intelligence would progressively reduce the cost of economically useful cognition. If that happened, value would begin moving away from the model itself and towards the assets required to apply it: proprietary data, distribution, trusted relationships, embedded workflows and physical infrastructure.

You might also like reading the Abundance Shock:

China’s latest advances do not prove that framework. They are, however, producing several of the outcomes it anticipated.

Models are improving faster than their prices are rising. Businesses can increasingly match different systems to different tasks instead of relying on one provider.

Attention is shifting away from the advertised cost of a token and towards the total cost of completing useful work. Capabilities previously limited to the largest and best-capitalised organisations are becoming accessible to a much broader market.

The economics should follow.
This transition also explains the progression of our own research.

During the first phase of the AI boom, value accrued to the bottlenecks that could not be expanded quickly enough. Advanced processors and high-bandwidth memory were scarce, and our South Korean equity exposure belonged to that phase.

EWY offered underappreciated exposure to an emerging memory shortage before the importance of memory to AI infrastructure became central to the market narrative.

Recognition eventually arrived. Earnings accelerated, South Korean equities appreciated dramatically and an underappreciated expression of the AI buildout became a consensus winner. Tier One harvested the gains and ultimately closed the position... not because the AI opportunity had ended, but because recognition had changed the asymmetry.

We were already turning towards the application layer. Not because physical infrastructure had ceased to matter, but because the next phase of value creation was likely to emerge somewhere less visible.

The first phase possessed a direct earnings bridge. Hyperscalers increased capital expenditure → semiconductor, memory and infrastructure suppliers converted those cheques into revenue. The application layer depended on a different mechanism: lower intelligence costs had to stimulate enough new usage to transform the economics of the companies applying it.

That mechanism is becoming easier to see.

Goldman Sachs estimates that consumer and enterprise agents could drive global token consumption to more than 24 times today’s estimated capacity by 2030.

Most of the projected growth comes not from conventional AI workloads but from agents performing longer, more frequent and increasingly autonomous sequences of work. The exact magnitude is necessarily uncertain. Adoption could proceed more slowly, model efficiency could improve faster than expected and many proposed applications may never become economical.

The composition of the estimate is more important than the precise total.

Non-agent workloads contribute only a small share of the expected increase. Most of the growth comes from consumer and enterprise applications: agents searching, organising and transacting for individuals, and systems operating across software development, customer service, research, administration and commercial workflows.

That is the application-layer thesis in one chart.

As intelligence becomes cheaper, it can be deployed more frequently, across a wider range of problems and at progressively lower levels of economic value. The cost of each task may fall, yet the number of tasks attempted can rise much faster.

Abundance does not merely reduce expenditure.

It can create demand.

China’s latest advances strengthen that possibility. More capable open-weight models, lower task-completion costs and a broader competitive ecosystem should make agentic applications cheaper to build, easier to deploy and less dependent on any single provider.

The result could be an economy that consumes substantially more intelligence even as the cost of each individual unit declines.

That has important consequences for market leadership.

For most of the recent cycle, the companies receiving AI capital-expenditure cheques have materially outperformed those paying them. Hardware suppliers enjoyed immediate demand, expanding margins and visible order books. The platforms absorbed enormous expenditure while the return on that investment remained difficult to measure.

We do not believe that the hardware opportunity has disappeared. If token consumption develops remotely along the path Goldman Sachs envisages, the world will continue to require substantial investment in processors, memory, networking and data centres.

But the distribution of returns can still change.

China’s progress challenges the assumption that the economics of AI will remain concentrated around permanently scarce and increasingly expensive computation. If equivalent or superior output can be produced at a lower cost, the scarcity premium enjoyed by parts of the hardware stack becomes harder to sustain.

At the same time, the platforms financing that infrastructure gain access to cheaper models and a larger opportunity to monetise the capacity they have already built.

The first phase rewarded the recipients of AI spending.

The next may increasingly reward its payers... and, beyond them, the companies capable of converting cheaper intelligence into better products, lower costs and stronger customer relationships.

China’s latest breakthroughs could become an important catalyst for that rotation.

They provide further evidence that capable intelligence is becoming more abundant, that application economics are improving and that the beneficiaries of the AI boom may not remain the same as its earliest winners.

That expectation has significant implications for the Alpha Tier Model Portfolio. Several positions already sit on the favourable side of this transition. The latest evidence strengthens the case for them, although it does not necessarily mean that the next unit of capital should be allocated to the same theme.

That is where we turn next.

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the July 2026 issue of Alpha Tier. The research, views and Alpha Tier Model Portfolio information are stated as of 24 July 2026, 4:00 p.m. Eastern Daylight Time. The original issue was first disseminated to paid subscribers on 24 July 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. The Alpha Tier Model Portfolio is an illustrative research portfolio and does not represent client assets, transactions executed by VMF Research or the performance of an investable fund or managed account.

The analysis combines thematic, fundamental, technological, portfolio-construction and market research within a medium- to long-term investment framework. Statements concerning the development of artificial intelligence, the economics of Chinese models, future token consumption and the possible migration of value towards the application layer are analytical judgments and conditional hypotheses, not assurances. Sources and the basis for material factual claims are identified throughout the article. Views and Model Portfolio conclusions may change as technological, competitive, regulatory, geopolitical, corporate or market evidence evolves and are reviewed through VMF Research’s monthly publications, with weekly or ad hoc updates where appropriate.

Investments in China, emerging markets, technology companies and AI-related financial instruments involve substantial risks, including regulatory intervention, geopolitical tension, restrictions on capital or technology flows, currency movements, governance concerns, market-access limitations, intense competition, pricing pressure, technological obsolescence, capital misallocation, valuation compression, liquidity constraints and the possible loss of capital. Improvements in model capability or lower intelligence costs may not produce higher corporate earnings, successful commercial adoption or superior investment returns. Technical and relative-strength signals may also fail.

Any references to historical recommendations or Model Portfolio performance relate to published reference prices and illustrative research portfolios. They do not represent returns earned by clients and should not be interpreted as a complete measure of investment performance.

Disclosure of interests: as of the research cut-off, neither VMF Research, Lda. nor Vasco Marques de Freitas personally held any financial instrument included in the Alpha Tier Model Portfolio. A legal entity controlled by Vasco Marques de Freitas held a long position in the KraneShares CSI China Internet ETF (KWEB), which provides exposure to several Chinese internet and application-layer companies relevant to the themes discussed in this article. This holding constitutes a financial interest and potential conflict of interest and should be considered when evaluating the analysis. Its existence does not validate the conclusions, reduce the risks described or imply that any financial instrument is suitable for a particular reader.

Neither VMF Research nor the author received compensation from any issuer discussed in connection with the preparation of this research, and no issuer reviewed, approved or amended its conclusions before first dissemination.

Past performance is not indicative of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

Discovery captures the imagination.

Royalties capture the cash flow.

Our previous article, The Discovery Machine, introduced the first of two biotechnology companies added to the Quality Model Portfolio in July’s issue of VMF’s Security Selection.

Regeneron represented the scientific engine: proprietary human data, experimental infrastructure and the ability to turn biological evidence into medicine. Cheaper intelligence may make that system more productive, but the company must still identify the right target, develop the treatment and survive the clinical process.

The second company offers a very different route into the same opportunity.

Halozyme does not need to discover the next blockbuster.

It gets paid when somebody else already has.

Its core technology allows certain medicines that would ordinarily require intravenous infusion to be administered beneath the skin, often in a fraction of the time. Pharmaceutical partners remain responsible for development, regulatory approval and commercialisation. Halozyme provides the delivery layer and receives fees, enzyme sales and, most importantly, royalties linked to partner revenues.

The attraction lies in the separation between participation and responsibility.

Halozyme participates in the success of multiple medicines without financing the full cost of creating each one. Its partners carry most of the scientific, clinical and commercial burden. Halozyme collects a percentage when the products sell.

That economic architecture should be familiar to long-time VMF Research readers.

We admire royalty businesses, and several appear across our Model Portfolios. The best combine exposure to valuable assets with limited requirements for incremental capital. A mineral royalty company does not need to construct and operate every mine. A music-rights company can continue collecting from a successful catalogue without financing every new recording. A biotechnology royalty company does not need to discover and commercialise every medicine on which it earns.

The assets differ. The principle does not.

Allow somebody else to commit much of the capital and assume much of the execution risk, while retaining a contractual claim on the upside.

Halozyme’s financial statements are beginning to reveal the power of that model. Royalty revenue is growing faster than the wider business, cash conversion is unusually strong and each additional dollar of partner sales requires relatively little incremental expenditure. The company is not merely expanding. Its revenue mix is becoming more valuable as it grows.

The risks remain real. Halozyme depends on medicines and commercial decisions it does not control. Its intellectual property has finite lives, competing delivery technologies exist, and its balance sheet carries more leverage than we would ideally prefer.

Royalty economics reduce certain risks.

They do not eliminate them...

The July issue was first published for paid subscribers on 17 July 2026. At the time of writing on 29 July, both biotechnology additions were trading above their initial reference prices. That early response is encouraging, although a matter of days cannot validate a medium- to long-term thesis.

Together, the two setups express the same biotech opportunity through complementary business models.

Regeneron must improve the probability of discovery.

Halozyme participates after much of the hardest scientific work has already succeeded.

The previous article introduced the discovery machine.

This one introduces the tollbooth.

Good reading.

Regeneron must discover the next medicine.

Halozyme ( ) gets paid when somebody else already has.

Its core ENHANZE technology uses a patented recombinant enzyme, rHuPH20, to temporarily break down hyaluronan beneath the skin. That creates space for large volume medicines that would otherwise require intravenous infusion to be administered subcutaneously, often more quickly, conveniently and potentially outside a hospital setting.

Halozyme licenses the technology to pharmaceutical partners, which remain responsible for developing, obtaining approval for and commercialising the resulting medicines. In return, Halozyme receives upfront fees, milestones, sales of the enzyme and, most importantly, royalties averaging a mid-single-digit percentage of partner sales.

The parallel with Altius Minerals is difficult to miss.

You might like also like reading

Altius does not need to build and operate every mine on which it owns a royalty. Its partners commit the capital, manage the assets and assume most of the operational risk. Altius participates when production succeeds.

Halozyme applies a similar economic architecture to biotechnology. Its partners bear most of the clinical, regulatory and commercial burden, while Halozyme collects a percentage of sales from medicines whose delivery its technology improves.

The underlying assets are different (mineral rights in one case, intellectual property and licensing agreements in the other) but the attraction is the same:

Participate in the upside of multiple assets without financing the full cost of creating each one.

That model is becoming increasingly visible in the financial statements.

Between 2021 and 2025, Halozyme’s revenue rose from $443 million to almost $1.4 billion. Royalty revenue increased even faster, from $204 million to $868 million, expanding from 46% to more than 62% of the total. Free cash flow rose from approximately $298 million to $645 million, equivalent to 46% of 2025 revenue. First-quarter 2026 revenue increased 42% to $377 million, while royalty revenue rose 43% to $241 million.

These are unusually attractive economics.

As partner medicines grow, Halozyme collects additional royalties without carrying a proportionate increase in research, manufacturing or commercial expenditure. The principal current drivers include subcutaneous versions of DARZALEX, Phesgo and VYVGART, while recently signed agreements with GSK, Vertex and Oruka expand the opportunity across ENHANZE and the newer Hypercon platform.

Management expects 2026 revenue of $1.71–$1.81 billion, royalty revenue of $1.13– $1.17 billion and adjusted EBITDA of $1.13–$1.21 billion.

Halozyme is therefore not merely growing.

Its revenue mix is improving as it grows.

Royalties are replacing less valuable and less predictable sources of revenue, expanding cash conversion and reducing the amount of incremental capital required to support each additional dollar of sales. The company has used that cash aggressively on behalf of shareholders. Diluted shares declined from approximately 147 million in 2021 to 124 million in 2025.

Halozyme repurchased 19.1 million shares under its previous programme at an average price of $39.31, followed by approximately seven million additional shares during 2025 at an average price close to $52.

The board has now authorised another $1 billion of repurchases through 2028, with at least $400 million expected during 2026.

At the approximately $77.50 price shown in the chart above, the shares trade near ten times the midpoint of management’s 2026 non-GAAP earnings guidance. That is an undemanding valuation for a company growing royalty revenue above 30%, converting a large proportion of revenue into cash and reducing its share count materially.

But the model is not without vulnerabilities.

Halozyme still depends on the success of medicines it does not control.

Partners determine development priorities, clinical investment, pricing and commercial execution. A trial failure, weak launch or strategic decision to abandon a programme can eliminate an expected royalty stream without Halozyme being able to change the outcome.

Revenue also remains concentrated in several important partner products, while alternative subcutaneous-delivery technologies, including Alteogen’s competing hyaluronidase platform, could weaken Halozyme’s negotiating power or reduce future adoption. ENHANZE patent expirations and contractual royalty reductions must also be considered product by product and country by country.

The balance sheet is the element we like least.

At the end of March, Halozyme carried approximately $2.14 billion of debt, against only $319 million of cash and marketable securities, leaving net debt close to $1.83 billion.

Much of that debt consists of low-coupon convertible notes maturing in 2031 and 2032, and the company’s powerful cash generation makes the burden manageable.

At the midpoint of 2026 adjusted EBITDA guidance, net leverage would fall towards 1.6 times. Nevertheless, acquisitions, continued repurchases and financial leverage now compete for the same cash.

The company has less room for strategic error than the income statement alone might suggest.

The bear case is therefore straightforward.

Partner-product growth could slow, competition could erode the value of ENHANZE,

newer platforms may fail to reproduce its success, and management could continue deploying borrowed capital into acquisitions or repurchases at unattractive prices. Halozyme’s economic model substantially reduces individual drug-development risk; it does not eliminate partner concentration, intellectual-property risk or capital allocation risk.

We believe those risks are adequately compensated.

Management’s 2026 outlook makes the strength of the underlying model unusually clear. Total revenue is expected to reach $1.71–$1.81 billion, implying growth of 22%–30%, while royalty revenue is projected to increase 30%–35% to as much as $1.17 billion. More strikingly, adjusted EBITDA is guided to $1.13–$1.21 billion, an increase of 71%–83%, with non-GAAP diluted earnings per share expected to rise 87%–99%.

Those figures require careful interpretation. Management’s adjusted EBITDA excludes several items captured in the GAAP financials, including acquisition related charges and intangible amortisation, and is therefore not directly comparable with the GAAP EBITDA published before.

We prefer to underwrite the business using the more conservative measure and treat the company’s guidance as evidence of operating momentum rather than a substitute for our own valuation work.

Even on that basis, the outlook is powerful.

Halozyme’s highest-quality revenue stream is becoming a larger share of the business, partner medicines continue to expand, and the incremental economics remain exceptionally attractive. The company does not need to identify the next blockbuster itself. It needs a growing portfolio of established partners to keep selling medicines that physicians and patients already value, while Halozyme collects a royalty each time they do.

The long-term chart reflects that quality.

For more than two decades, the shares have advanced within a remarkable rising channel, repeatedly absorbing volatility without breaking the broader trend. Price remains above rising long-term moving averages, while relative strength against the wider equity market remains firmly positive. We see no reason to conclude that this pattern has reached its end.

Against the backdrop of the healthcare and biotechnology opportunity we have developed across VMF Research, Halozyme represents a compelling addition to our roster: a capital-light royalty engine, growing profitability, powerful cash conversion and exposure to multiple successful medicines without assuming the full clinical and commercial risk of creating each one.

Regeneron gives us ownership of the discovery machine.
Halozyme gives us the tollbooth.

You might also like reading, The Intelligence Dividend - How cheaper AI cuts biopharma’s $300B failure risk and drives the next healthcare rotation:

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the July 2026 issue of VMF’s Security Selection. The research, views, valuation analysis and Quality Model Portfolio information are stated as of 17 July 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 17 July 2026 at 9:00 p.m. Eastern Daylight Time. Any reference to subsequent price performance is stated separately and does not amend the original analysis.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. VMF Research’s Model Portfolios are illustrative research portfolios and do not represent client assets or transactions executed by VMF Research.

The analysis combines thematic, fundamental, scientific, valuation and market research within a medium- to long-term investment framework. Forecasts, valuation assessments and statements concerning future revenue, royalties, cash generation or operating leverage are analytical judgments rather than assurances. Sources and the basis for material factual claims are identified throughout the article. Views and conclusions may change as scientific, clinical, regulatory, competitive, corporate or market evidence evolves and are reviewed through VMF Research’s monthly publications, with weekly or ad hoc updates where appropriate.

Investments in Halozyme Therapeutics and other biotechnology companies involve substantial risks, including dependence on partner products and commercial decisions, clinical or regulatory setbacks affecting licensed medicines, product and customer concentration, loss or expiry of intellectual-property protection, competing drug-delivery technologies, pricing and reimbursement pressure, leverage, acquisition and capital-allocation risk, market volatility and the possible loss of capital. A royalty-based business model may reduce some development and commercial risks without eliminating them. Historical growth, cash conversion and technical trends may not persist, and valuation assumptions may fail.

Disclosure of interests: as of the research cut-off, neither VMF Research, Lda., Vasco Marques de Freitas personally, nor any legal entity controlled by him held a position in Halozyme Therapeutics or Regeneron Pharmaceuticals. A legal entity controlled by Vasco Marques de Freitas held a long position in Altius Minerals Corporation, which is referenced in the article as a comparative example of a royalty-based business model. This constitutes a financial interest and potential conflict of interest and should be considered when evaluating the comparison. Neither VMF Research nor the author received compensation from any issuer discussed in connection with the preparation of this research, and no issuer reviewed or approved its conclusions before first dissemination.

Past performance is not indicative of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial intermediary or adviser before making an investment decision.

The most valuable AI asset in biotechnology cannot be downloaded.

It has to be found in millions of human bodies, validated in laboratories and proved in patients.

That distinction sits at the centre of the research we have been publishing throughout July.

In The China AI Reset, we examined how improving Chinese models are driving down the cost of capable intelligence. In The Intelligence Dividend, we followed that supply shock into healthcare, an industry where better decisions can save years of work and hundreds of millions of dollars. Most recently, When Cheap Starts to Lead showed that the opportunity is no longer supported by valuation alone. Healthcare earnings expectations are improving, participation is widening and biotechnology has begun to behave like an emerging market leader.

The next step is the one VMF’s Security Selection was created to perform.

Move from the sector to the business.

More capable algorithms will eventually become available to every large pharmaceutical company. The same cannot be said for proprietary human data, experimental infrastructure or the ability to turn a biological insight into an approved medicine.

Regeneron has spent decades assembling that chain.

Its Genetics Centre has already sequenced more than three million samples and connected genetic information with clinical, proteomic and other molecular data. Additional agreements could expand the evidence base dramatically. The advantage lies not only in the size of the dataset, but in what surrounds it: human genetics identifies a possibility, laboratory biology tests the mechanism, proprietary technologies design the treatment and clinical development determines whether it works.

Artificial intelligence can make every stage more productive.

It can search relationships across genetic variants, proteins, diagnoses and treatment outcomes at a scale conventional research cannot match. It may improve target selection, patient stratification and trial design. It may help scientists recognise promising signals earlier.

Its greatest contribution could be more prosaic.

It may help Regeneron abandon the wrong programmes sooner.

That matters enormously when annual research spending exceeds $6 billion. A weak target killed one year earlier can create substantial value without producing a single new medicine. In biotechnology, avoiding an expensive failure can be almost as important as accelerating a success.

The share price tells a less ambitious story.

Investors see an ageing ophthalmology blockbuster, rising biosimilar competition, slowing reported growth and an increasingly expensive research organisation. Those concerns are real. EYLEA remains under pressure, Dupixent must continue carrying more of the economic burden, and biology has already demonstrated that even excellent science can fail in a pivotal trial.

That tension is what makes the company interesting.

Regeneron is being judged increasingly through the decline of the medicine that financed its expansion. Our research asks whether the more appropriate unit of analysis is the scientific institution built with those profits: a company with proprietary biological evidence, integrated discovery capabilities, substantial current cash flow and the financial strength to keep compounding its research advantage.

This is not an argument that AI will make biology predictable.

Biology rarely grants investors that luxury.

It is an argument that cheaper intelligence may increase the value of scarce evidence, improve the allocation of research capital and strengthen the economics of the organisations best equipped to convert data into medicine.

That is how the work fits together across VMF Research.

VMF’s Strategic Asset Allocation identifies where economic value and market leadership may be migrating. VMF’s Security Selection underwrites the individual companies positioned to capture that change. Alpha Tier considers how quality, asymmetry, conviction and risk budget should coexist inside a more concentrated Model Portfolio.

The research below was first published for paid subscribers on 17 July 2026 at 9:00 p.m. Eastern Daylight Time. Regeneron was one of two new additions to the Quality Model Portfolio in July’s issue of VMF’s Security Selection.

Later this week, we will publish the second.

Regeneron must discover the next medicine.

The other company gets paid when somebody else already has.

One owns the discovery machine.

The other owns the tollbooth.

Today, we open the first.

-

Good reading.

The most valuable dataset in artificial intelligence may not be written in words.

It may be written in the differences between millions of human bodies: why one person develops a disease while another remains protected; why a medicine works in one patient and fails in the next; which molecular changes precede illness; and which merely accompany it.

Regeneron ( ) has spent years assembling that evidence.

Its Genetics Centre has already sequenced more than three million samples and connected genetic information with clinical, proteomic and other molecular data. A separate agreement with Truveta could extend the work to as many as ten million consented participants. More recently, Regeneron secured access to de-identified health records covering approximately 300 million people through TriNetX, with the explicit objective of linking those clinical histories to its molecular datasets and using the resulting information in drug discovery, product development and the training of AI systems.

That is what makes Regeneron particularly relevant to the thesis we developed last week in VMF’s Strategic Asset Allocation.

The model is becoming cheaper.
The evidence required to make it useful is not.

Every large pharmaceutical company will gain access to more capable algorithms.

Far fewer possess proprietary human data at sufficient scale, the experimental infrastructure required to validate what those algorithms discover and the ability to transform the result into an approved medicine.

Regeneron owns all three.

It is a fully integrated biotechnology company: it identifies biological targets, engineers potential treatments, conducts clinical trials, manufactures medicines and commercialises them directly or through selected partners. Its internal research has produced 15 approved or authorised medicines over roughly fifteen years, while nearly 50 candidates are now moving through clinical development across six therapeutic areas. Most originated inside Regeneron’s own laboratories.

Yet the market is increasingly valuing the company through a different lens.

It sees the decline of EYLEA, the ophthalmology medicine that financed much of Regeneron’s expansion. That concern is legitimate. The original product is losing share to competition and biosimilars, while EYLEA HD has not yet grown quickly enough to stabilise the combined franchise.

Our thesis is not that the market has misunderstood the EYLEA problem.
It is that it may be using the wrong unit of analysis.

Regeneron is being treated increasingly as the owner of an ageing blockbuster.

We believe it should be valued as a scientific institution that has repeatedly created blockbusters... and whose proprietary data may become more valuable as the cost of intelligence falls. That distinction is the investment opportunity.

From Biological Evidence to Medicine

Most drug discovery begins with a hypothesis (the same starting point as any rigorous investment process).

Regeneron is attempting to begin with human evidence. Its Genetics Centre searches for naturally occurring genetic variations associated with disease, protection or treatment response. A harmful mutation may reveal a biological pathway that should be blocked. A protective mutation can be even more valuable: it may show that reducing the activity of a particular gene is both beneficial and tolerated by the human body.

That does not make the resulting target automatically investable. Correlation is not causation, and a statistically compelling relationship may still prove impossible to manipulate safely.

Regeneron’s advantage is that the dataset sits inside a wider experimental system. The company can test genetic associations using organoids, RNA interference, CRISPR models and humanised mice. Its proprietary VelociSuite technologies can then generate fully human antibodies, bispecific antibodies, T-cell receptors and other therapeutic formats designed to act on the validated target. Regeneron can subsequently carry the candidate through clinical development, manufacturing and commercialisation.

The economic chain is therefore unusually complete:

Human genetics identifies the possibility. Biology tests the mechanism. Technology designs the medicine. Clinical evidence determines whether it works.

Artificial intelligence can improve several decisions inside that chain. It can search relationships across genetic variants, proteins, diagnoses, treatments and longitudinal outcomes at a scale beyond conventional human analysis. It may help identify causal pathways, prioritise targets, discover biomarkers, select the patients most likely to respond and design trials capable of producing clearer evidence.

But the most valuable contribution may be less glamorous.

AI may help Regeneron decide what not to develop.

The company expects to spend between $6.45 billion and $6.68 billion on research and development during 2026. When the annual research budget is that large, terminating an unpromising programme one year earlier can create substantial value... even if the technology never discovers a blockbuster by itself.

That is how we think investors should frame the opportunity.

The near-term AI benefit is unlikely to appear as a separate revenue line. It may first appear through better capital allocation inside the laboratory: fewer weak targets reaching expensive trials, more precise patient selection and a higher proportion of resources directed towards programmes with strong human evidence.

A modest improvement in those decisions would matter.

A revolutionary improvement is not required.

The Current Business Still Has to Carry the Thesis

Regeneron’s AI and data advantage is compelling, but it cannot substitute for the economics of the existing business.

Those economics are strong...

The five-year financial record tells a more nuanced story than the share-price decline suggests. Reported revenue fell sharply from $16.1 billion in 2021 to $12.2 billion in 2022, before recovering to $13.1 billion in 2023, $14.2 billion in 2024 and $14.3 billion last year. The recovery has therefore been persistent, but modest: revenue has grown at roughly 5.6% annually since 2022, with growth slowing to only 1% in 2025.

The original decline was less alarming than it appears.

Approximately $6.2 billion of Regeneron’s 2021 revenue came from REGEN-COV, its Covid-19 antibody treatment. The emergence of Omicron rendered the medicine largely ineffective, causing the FDA to restrict its use and eliminating what had always been an exceptional source of revenue. Excluding Regeneron’s Covid antibodies, underlying revenue increased 17% in 2022.

The apparent collapse was therefore primarily the disappearance of a pandemic windfall... not the deterioration of the core business.

What followed, however, has been less impressive.

Profitable Enough to Keep Investing

Regeneron remains extraordinarily well financed.

At the end of March, the company held approximately $18.5 billion of cash and marketable securities, against less than $2 billion of long-term debt and approximately $720 million of finance-lease liabilities. Net financial resources consequently approached $15.8 billion, equivalent to almost one-quarter of the company’s current market capitalisation ($73 billion).

The operating business also continues to generate substantial cash. In 2025, Regeneron produced approximately $5.0 billion of operating cash flow. After nearly $900 million of capital expenditure, free cash flow reached roughly $4.1 billion, equivalent to almost 29% of revenue. The company used that capacity to repurchase $3.4 billion of shares, pay its first $370 million of dividends and continue expanding its scientific and manufacturing infrastructure.

It did so while spending $5.85 billion on research and development, or almost 41% of revenue. During the first quarter of 2026, R&D rose another 16% to $1.54 billion (nearly 43% of quarterly revenue).

That combination is rare:

Regeneron can invest more than $6 billion annually in future medicines while continuing to generate billions in free cash flow and maintaining a fortress balance sheet.

The financial capacity is not in question.
The productivity of the capital being deployed is.

Why Revenue Is Languishing

Regeneron’s present revenue profile is being pulled in opposite directions. Net product sales fell from $7.6 billion in 2024 to $6.3 billion in 2025. Over the same period, collaboration revenue increased from $6.1 billion to $7.3 billion. Most of that increase came from Regeneron’s share of the rapidly expanding profits generated by Dupixent, while most of the pressure came from the declining EYLEA ophthalmology franchise.

In simple terms: Dupixent is replacing what EYLEA is losing but not yet creating enough incremental growth to make the broader company accelerate.

The accounting also obscures some of the underlying progress. Sanofi records Dupixent’s global product sales. Regeneron reports its contractual share of the collaboration profits rather than the entire revenue generated by the medicine. Dupixent and Kevzara sales reached $18.4 billion in 2025, up from $14.6 billion, while Regeneron’s Sanofi collaboration revenue increased by approximately $1.35 billion.

The growth is real.
It simply appears differently in Regeneron’s financial statements.

EYLEA is the mirror image. Regeneron records US sales directly, so the decline is fully visible in reported product revenue. Competitive treatments, falling net prices, the migration of patients to EYLEA HD and the arrival of biosimilars have all pressured the original formulation. More biosimilar launches are expected during the second half of 2026.

The company is therefore undergoing a difficult but understandable transition:

• From pandemic revenue to recurring franchises.

• From legacy EYLEA towards EYLEA HD.

• From direct product revenue towards greater Dupixent collaboration economics.

• From a smaller research organisation towards a much broader (and more expensive) development platform.

The business is not stagnant.
Its strongest growth engine is being offset by its most visible decline.

Where the Next Revenue Can Come From

The bull case does not depend on one heroic breakthrough. It depends on several existing and emerging engines beginning to work at the same time.

Dupixent remains the clearest source of growth.

Its underlying biology extends across multiple Type 2 inflammatory diseases, giving Regeneron and Sanofi room to expand the franchise through new indications, geographies and age groups. Regeneron’s share of commercial profits rose strongly again in the first quarter, while the remaining historical development balance owed to Sanofi is nearing repayment. Once cleared, a larger share of future collaboration economics should flow through to Regeneron. Dupixent will inevitably slow as the base becomes larger, but it does not need to sustain 30%-plus growth. Continued double-digit expansion would be enough to support the company while the rest of the portfolio matures.

Libtayo is also becoming economically relevant. First-quarter sales rose 54% to $438 million, with Regeneron retaining greater control over the economics than it does with Dupixent. Lynozyfic broadens the oncology portfolio, while garetosmab and cemdisiran could receive regulatory decisions before year-end. None is likely to replace EYLEA alone. Together, they reduce the need for any single asset to do so.

The longer-duration opportunity lies in Factor XI inhibition, metabolic disease, oncology, genetic medicines and additional immunology programmes. This is where Regeneron’s data and AI advantage matters most. The immediate benefit is unlikely to appear as a separate revenue line. It should appear first in better decisions over which programmes deserve capital (and which should be abandoned earlier). When annual R&D spending exceeds $6 billion, avoiding one costly failure can be almost as valuable as accelerating one success.

EYLEA HD is the bridge between the current business and that future.

The product is gaining traction. As the chart on page 10 makes clear, EYLEA HD is gaining share within the franchise, but not yet quickly enough to offset the decline of legacy EYLEA. Some of the recent weakness reflects seasonality and inventory normalisation... but the more important pressure is structural, driven by lower pricing, competing therapies and biosimilars.

EYLEA HD nevertheless has credible advantages: longer and more flexible dosing intervals, an expanded label, a potential pre-filled syringe and deep relationships with retinal specialists. Those strengths may eventually return the franchise to modest growth.

But that is not our base case.

The hurdle is lower... and more realistic:

EYLEA HD does not need to recreate peak EYLEA. It needs to slow the decline enough for Dupixent, Libtayo and the pipeline to become visible through it.

A return to growth would be meaningful upside. Stability would already be enough to support the thesis.

What Can Go Wrong

The principal risk is that Regeneron’s transition takes longer than the market or the financial model can tolerate.

Dupixent and the US EYLEA franchise together still accounted for approximately 71% of first-quarter revenue. Diversification across nearly 50 clinical programmes is not the same as diversification of present cash flow.

• Several developments could weaken the thesis materially:

• EYLEA biosimilars could accelerate pricing and volume erosion before EYLEA HD reaches sufficient scale.

• Dupixent growth could normalise more quickly than expected, leaving the company dependent on products that are not yet economically material.

• The pipeline could produce medically useful but commercially modest medicines.

• R&D expenditure could continue rising while operating margins compress.

• The company’s large securities portfolio can make GAAP net income more volatile and less representative of underlying operating earnings.

• AI may improve discovery efficiency without solving the late-stage clinical failures that consume the greatest amount of capital.

• Pharmaceutical pricing policy could constrain returns on future innovation.

The fianlimab1 melanoma failure is a useful reminder. Regeneron can possess world class scientists, proprietary data and sophisticated technology—and biology can still refuse to cooperate.

The quality of the platform reduces risk.

It does not remove it.

Valuation: What Has to Happen by 2028

Regeneron is not priced as a distressed biotechnology company. Nor does the current valuation require flawless execution.

At approximately $678 per share, the equity is valued at roughly $73.0 billion. Against trailing sales of approximately $14.9 billion, free cash flow of $4.1 billion and EBITDA of about $4.2 billion, the shares trade at 17.7 times free cash flow. The company’s $15.8 billion net-cash position reduces enterprise value to approximately $57.1 billion, equivalent to 3.8 times sales and 13.7 times EBITDA.

Those multiples are not obviously cheap in isolation. They become more interesting when attached to a net-cash company that continues to generate substantial free cash flow while investing more than 40% of revenue in research.

We value Regeneron primarily on 2028 free cash flow. That gives Dupixent time to expand, EYLEA HD time to moderate ophthalmology erosion and the emerging portfolio time to become more economically visible. Importantly, our base case assigns no explicit revenue to AI. Better intelligence is reflected only through the possibility of more productive research allocation and greater durability beyond the forecast period.

The bear case does not require a scientific collapse. EYLEA erosion simply remains severe, Dupixent slows before new products become material and elevated R&D spending prevents margins from recovering. Regeneron remains profitable, but the market values it increasingly like a mature pharmaceutical company.

The base case assumes something less dramatic: Dupixent continues growing at a sustainable rate, EYLEA HD bends the ophthalmology decline, Libtayo and the emerging portfolio broaden the revenue base and free-cash-flow margins remain close to recent levels. A modest rerating to 19 times FCF would produce approximately 33% market-cap appreciation.

The bull case requires the engines to align. Ophthalmology stabilises, Dupixent remains unusually durable, several new medicines become commercially meaningful and investors begin assigning greater value to Regeneron’s scientific platform and proprietary data. Under those conditions, double-digit sales growth and operating leverage could nearly double the company’s equity value.

Applying probabilities of 20% to the bear case, 55% to the base case and 25% to the bull case produces a probability-weighted market capitalisation of approximately $99 billion, or roughly 35% above the current level, before dividends and share repurchases.

That is attractive, but not riskless.

The valuation offers a reasonable price for exceptional scientific and financial quality... not a margin of safety large enough to ignore product concentration or clinical attrition. It supports a 2% position, while preserving room to increase the allocation if the financial and technical evidence continues to improve.

The Chart Is Reaching a Decision

The technical setup is becoming more constructive, but confirmation has not yet arrived.

After falling sharply from its 2024 peak, Regeneron has spent more than a year compressing inside a large triangle. Descending resistance from the previous high is converging with rising support from the 2025 low, leaving the shares close to a visible decision point.

Price remains marginally below the 20- and 50-week moving averages and well beneath the 200-week average. Those levels still represent resistance.

Encouragingly, however, the shares are challenging the shorter-term averages as relative strength against the broader equity market begins to improve after a prolonged period of underperformance.

The triangle can still resolve in either direction.

A breakout above descending resistance and the nearby moving averages would indicate that investors are beginning to look beyond EYLEA’s erosion and towards Dupixent’s durability, the pipeline and Regeneron’s longer-term scientific optionality. A break below rising support would undermine the recovery and expose meaningful downside.

Our constructive view on healthcare (and biotechnology in particular) raises the probability, in our judgement, of an upside resolution. Sector participation is broadening, biotech has already broken out and Regeneron is approaching resistance as its relative performance begins to repair.

This has been a lengthy thesis because the opportunity demands it. The upside is substantial, but so are the product-concentration, clinical and execution risks; the downside must be underwritten rather than dismissed. Our next recommendation will require less extensive treatment because, in our assessment, its fundamental risk profile is materially lower.

You might also like reading, The Intelligence Dividend - How cheaper AI cuts biopharma’s $300B failure risk and drives the next healthcare rotation:

Important Disclosure

This article contains general investment research produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. It reproduces an excerpt from the July 2026 issue of VMF’s Security Selection. The research, views, valuation analysis and Quality Model Portfolio information are stated as of 17 July 2026, 4:00 p.m. Eastern Daylight Time, unless another date is expressly identified. The original issue was first disseminated to paid subscribers on 17 July 2026 at 9:00 p.m. Eastern Daylight Time.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s individual objectives, financial circumstances, knowledge, experience, liquidity requirements or tolerance for risk. VMF Research’s Model Portfolios are illustrative research portfolios and do not represent client assets or transactions executed by VMF Research.

The analysis combines thematic, fundamental, scientific, valuation and market research within a medium- to long-term investment framework. Forecasts, scenario outcomes, valuation estimates and statements concerning the potential effects of artificial intelligence are analytical judgments rather than assurances. Sources and the basis for material factual claims are identified throughout the article. Views and conclusions may change as scientific, clinical, regulatory, corporate or market evidence evolves and are reviewed through VMF Research’s monthly publications, with weekly or ad hoc updates where appropriate.

Investments in Regeneron Pharmaceuticals and other biotechnology companies involve substantial risks, including clinical-trial failure, regulatory rejection, product concentration, loss of intellectual-property protection, biosimilar and competitive pressure, adverse pricing or reimbursement decisions, rising research expenditure, scientific uncertainty, market volatility and the possible loss of capital. Artificial intelligence may improve research productivity without producing successful medicines, higher earnings or superior investment returns. Technical patterns and valuation assumptions may also fail.

Disclosure of interests: as of the research cut-off, neither VMF Research, Lda., Vasco Marques de Freitas personally, nor any legal entity controlled by him held a position in Regeneron Pharmaceuticals or in any other financial instrument mentioned in this article. Neither VMF Research nor the author received compensation from any issuer discussed in connection with the preparation of this research, and no issuer reviewed or approved its conclusions before first dissemination.

Past performance is not indicative of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial adviser before making an investment decision.

This Weekly Update was first published for Tier One subscribers on 25 July 2026.

We argued that long-term Treasury yields had entered what we call the TACO Zone: the point at which tighter financial conditions begin imposing enough political and economic pain to make a Trump retreat increasingly likely.

Two days later, President Trump has paused US strikes on Iran. Oil is falling sharply, Treasury yields are easing and risk assets are rallying.

The TACO may now be unfolding in real time.
Here is what we wrote before the turn.

Long-term yields have pushed into the zone we label the TACO Zone.

The 30-year Treasury yield is testing and probing the 5.16–5.20 area after a sustained rise along its multi-year uptrend. We call it the TACO Zone for a reason. At these levels, higher long-end yields begin to impose visible political and economic costs: rising mortgage rates, pressure on equity valuations, and tighter financial conditions that become harder to ignore.

History suggests this is precisely the point at which incentives can shift. For an administration focused on growth and market stability, the political calculus of allowing yields to break decisively higher starts to deteriorate.

The temptation to deliver a TACO (Trump Always Chickens Out), to soften rhetoric, adjust policy signals, or create conditions that ease the long end, tends to rise once the pain threshold is reached.

A decisive break above the zone would tighten financial conditions further and raise the bar for our Dovish Shock confirmation ladder.

The market is still priced for a restrictive path; this is the latest reminder that monetary evidence must still improve before the thesis can move from alive to confirmed.

At the same time, the very existence of the TACO Zone highlights the growing tension between market pricing and political incentives.

Energy has been the clearest beneficiary of the renewed geopolitical risk premium. XLE has broken higher from a textbook flag consolidation, accompanied by improving relative strength versus the broader US equity market.

The existing allocation is responding exactly as the Iran related supply fears and oil volatility would suggest. Absolute momentum and relative performance are both moving in the same direction for the first time in several months.

At the same time, another important dynamic is gripping markets: growing uncertainty about the ultimate size, speed and profitability of the AI capex boom.

Recent breakthroughs from Chinese laboratories continue to change the calculation, both the size of the overall pie and how the slices may ultimately be distributed. Capability is improving while costs compress.

That combination pressures the scarcity premiums that defined the first phase of the trade.

Our recent exit from the remaining EWY position after a 238% total return therefore appears increasingly timely.

We entered when memory was still an underappreciated bottleneck. We left once the market had fully discovered the thesis and price had begun to embed an unusually optimistic future. The subsequent reversal in South Korean equities simply reinforces the value of harvesting when recognition arrives.

No change to the Model Portfolio is required.

The existing exposures already capture the energy risk premium and the downstream application-layer opportunity. Higher long-end yields remain a headwind that the Dovish Shock thesis must still overcome, even as political incentives to deliver a TACO begin to build. We will continue to measure price against process rather than force activity for its own sake.

Good investing!

Vasco Marques de Freitas, CFA, CMT

Important Disclosure

This article reproduces general investment research first published for Tier One subscribers on 25 July 2026. It was produced by Vasco Marques de Freitas, CFA, CMT, Founder and CEO of VMF Research, Lda. The original analysis, charts and conclusions reflect the information available at the time of publication and have not been revised with hindsight. Any introductory comments concerning subsequent events were added when the article was republished and should be distinguished from the original research.

The publication contains information recommending or suggesting an investment strategy. It is not personalised investment advice and does not consider any reader’s objectives, financial circumstances, knowledge, experience or tolerance for risk. The Tier One Model Portfolio is an illustrative research portfolio and does not represent client assets or transactions executed by VMF Research.

The analysis combines macroeconomic, political, fundamental and technical evidence within a medium- to long-term framework. The “TACO Zone” and “Dovish Shock” are conditional analytical hypotheses, not forecasts or assurances. Subsequent developments may appear consistent with the analysis without confirming its broader conclusions, and market conditions can reverse rapidly. Views are reviewed through VMF Research’s monthly publications and may be updated through weekly or ad hoc research.

Equities, energy-sector instruments, sovereign bonds and interest-rate-sensitive assets involve substantial market, geopolitical, policy, liquidity and valuation risks. Technical signals may fail, political decisions may change without warning and investors may incur material losses.

Disclosure of interests: as of the original publication date, neither VMF Research, Lda. nor Vasco Marques de Freitas personally held any financial instrument mentioned in this article.

Past performance is not indicative of future results. Readers should conduct their own analysis and, where appropriate, consult an authorised financial adviser before making an investment decision.