• AI spending boom is masking deeper economic weakness
• Asset owners thrive while rising costs crush ordinary households
• Extreme valuations depend on near-perfect long-term AI growth

The financial markets currently present a striking paradox.

On one hand, stocks are hovering near record highs. On the other, consumer sentiment has fallen to historic lows, as reflected in the University of Michigan sentiment index.

This divergence between Wall Street and Main Street is not difficult to explain once you look beneath the surface.

While economists may celebrate inflation "normalising" toward 2% to 3%, consumers are still absorbing a far harsher reality.

Everyday essentials remain 20% to 30% more expensive than just a few years ago.

Non-discretionary costs—healthcare, insurance, rent, and utilities—do not revert to prior levels. They accumulate. And with wages struggling to keep pace with cumulative inflation over the past six years, the result is a structurally weakened consumer base.

The graphic below is about five months old (where prices have increased further) but it paints a picture that few politicians would like to admit:

This friction has accelerated a K-shaped economic divide.

The upper half of the K consists of households that own financial and real assets—equities, property, and other inflation-linked stores of value. Their wealth compounds as asset prices rise, and higher interest rates increasingly reward their savings.

The lower half consists of households reliant on wages and debt-funded consumption. They face higher borrowing costs, rising living expenses, and limited asset exposure.

The result is not just inequality—it is divergence in economic velocity.

The two halves of the economy are no longer moving at the same speed.

A Highly Reflexive Market Environment

Under a traditional macroeconomic framework, weak consumer confidence combined with rising geopolitical risk should be a headwind for equities.

However, as our K-Shaped chart shows above – markets continue to push toward record highs.

To understand why this divergence persists despite macro fragility, we need to look at where the marginal dollar of capital is now flowing.

The answer is not consumption (which is typically 70% of GDP).

And it is not traditional corporate investment.

It is artificial intelligence (AI) infrastructure.

This shift has created a market regime where capital flows—rather than economic breadth—are doing the heavy lifting for asset prices.

Investors are aggressively pricing in the long-term earnings power of AI systems, driving upward revisions in future cash flow expectations and compressing risk premia.

For now, this AI-driven capital boom is overwhelming broader macroeconomic concerns.

But it also creates a second-order effect that is easy to miss.

When markets remain resilient in the face of geopolitical tension and strong economic strain, it changes behaviour at both the investor and policy level.

Political leaders face reduced market discipline. If equities continue rising through escalating tariffs, energy shocks, or geopolitical instability, markets are effectively signalling tolerance.

At the same time, investors begin to interpret rising asset prices as validation that underlying risks are contained.

This is the essence of reflexivity: markets do not merely reflect reality—they influence it.

And in this environment, strong markets can encourage greater risk-taking, which in turn increases underlying system fragility.

The perception of stability does not eliminate risk. It delays its resolution.

From Pixels to Silicon Infrastructure

For much of the past decade, capital was concentrated in software platforms and digital applications expected to dominate the AI revolution at the margins.

But a structural shift has now occurred.

Capital is migrating away from software and toward the physical backbone of AI: semiconductors, memory, data centres, networking infrastructure, and energy systems. In short, from "pixels" to silicon.

The large technology platforms are no longer simply capturing profits from AI expansion—they are financing its infrastructure at unprecedented scale.

Capital expenditure requirements are rising so quickly that even the highest-margin companies in history are being forced into aggressive reinvestment cycles.

This is a modern Red Queen dynamic: companies must run faster just to maintain their position.

Even cash-rich platforms are increasingly funnelling earnings into upstream suppliers, particularly semiconductor and infrastructure firms.

Alphabet"s $80 billion equity raise—despite extraordinary internal cash generation—highlights how capital-intensive this phase of the AI build-out has become.

History offers a useful lens here.

Early gains in technological revolutions tend to accrue to the narrative layer.

Eventually, however, they accrue to the infrastructure bottlenecks that control supply.

Silicon and memory—once considered cyclical commodity businesses—are now being treated as strategic infrastructure.

This shift helps explain why traditional valuation frameworks are beginning to look stretched.

Earnings expectations are being repeatedly revised upward, but largely on the assumption that infrastructure demand remains structurally elevated.

However, this is where capital-cycle discipline becomes important.

If one were intentionally designing a scenario most vulnerable to future disappointment, it would look exactly like this: assume demand remains permanently strong, capital flows remain rational, and supply constraints persist indefinitely.

History suggests otherwise.

Infrastructure booms almost always resolve in the same way: capacity eventually catches up with demand, and when it does, pricing power compresses rapidly.

The key question is whether AI infrastructure is exempt from this cycle—or simply experiencing a larger version of it.

My view is that it is not exempt.

Because when valuations only make sense under flawless long-term assumptions, the margin for error becomes extremely small.

The Reflexivity Loop

This brings us to the second-order effects of today"s market structure.

The combination of strong equity markets and weak underlying economic sentiment creates a psychological trap: the belief that markets now possess a permanent floor.

We saw similar thinking during the late stages of the dot-com cycle.

Today, a historically large share of household wealth is concentrated in equities, particularly among higher-income households who also dominate asset ownership (i.e., the upper K).

This creates an implicit constraint on policy: governments become increasingly reluctant to pursue actions that could destabilise financial markets.

But the feedback loop runs both ways.

If markets continue rising despite geopolitical tension, policymakers feel less pressure to moderate aggressive fiscal or geopolitical strategies.

Meanwhile, investors begin to interpret rising equity prices as confirmation that systemic risks are being successfully managed.

This is reflexivity in action:

• Strong markets encourage risk-taking.
• That risk-taking increases structural fragility beneath the surface.
• The system appears stable precisely when it is becoming more sensitive to disruption.

The loop then reinforces itself:

However, the perception of safety does not remove risk—it concentrates it.

The Disconnect with Valuations

I will conclude this post with a word on valuations.

After all, it is what you pay which will have the greatest bearing on your eventual returns.

Investor optimism today sits near extremes historically associated with major speculative periods.

The Shiller CAPE ratio — which smooths earnings over a decade — now sits above levels seen before the 1929 crash and approaches the highs reached during the dot-com bubble.

What"s more – the Shiller Excess CAPE Yield for the S&P 500 is now just 1.28

Every other time we have approached this ratio, subsequent returns over the coming decade have been flat to negative.

Don"t get me wrong – the expansion in asset prices can continue in the short-term (e.g., where short-term could be up to two to three years). I don"t pretend to know how long.

However, the longer-term view is not positive.

So why do investors firmly believe this time is different?

Primarily through near-term earnings growth.

If analysts are correct about the scale of future AI-driven profits, many current valuations can be defended.

But we should be careful….

The belief that historical valuation limits no longer apply is fraught with risk.

The weakness in that argument becomes clearer when looking beyond earnings margins and focusing on sales multiples instead.

Price-to-sales ratios across many areas of the market have reached unprecedented levels.

That matters because sales are much harder to financially engineer than earnings.

When markets trade at extremely elevated sales multiples, there is very little room for disappointment.

For example, what happens if:

• inflation remains structurally higher (e.g., 3% becomes the new target) due to energy disruptions or ongoing supply constraints?
• interest rates stay elevated longer than expected? or
• if AI services commoditize and margins shrink?

Historically, higher-rate environments compress valuation multiples rather than expand them.

My take is current prices increasingly rely on near-perfect execution (with very little going wrong).

When investors pay extreme premiums for future growth, they are often pulling forward returns that would otherwise have belonged to the next decade.

Putting it All Together

Navigating the next decade will require investors to reject many of the short-term narratives dominating market commentary today.

The current environment — where massive infrastructure spending and labour-market resilience are masking geopolitical risks and historically extreme valuations — is unlikely to represent a permanent economic regime.

Over time, financial gravity tends to reassert itself. Or as I like to say – everything reverts to the mean.

• Infrastructure shortages eventually become oversupply.
• Capital booms eventually create excess capacity.
• Valuation extremes eventually face pressure from interest rates, competition, or slowing growth.

None of this means the AI transformation is not real.

It clearly is.

I"ve worked in the field of AI the past decade.

This is the largest technological transformation since the internet. We are giving humans a new superpower.

That"s not a debate.

History repeatedly shows that even the most powerful technological revolutions still experience capital-cycle excesses along the way.

For long-term investors, the goal is not to avoid innovation. Embrace it.

It is to separate sustainable cash flows from temporary euphoria.

Because when the current market cushion eventually loses its elasticity, the greatest margin of safety will belong to investors who understood the difference between a durable structural shift and a cyclical infrastructure boom.

Regards,
Adrian Tout