The Great Rotation: Why Big Tech Is Weak While the Broader Market Stays Strong

Big Tech looks weak as hyperscalers shift to massive capex - yet the broader market stays resilient. Learn why capital is rotating, what’s bottlenecking AI, and how debt and valuation distortions matter.

carlos.artiles•Jun 29, 2026

This market is sending two completely different signals at the same time.

On one side, mega-cap tech looks stressed. Capital spending is exploding, margins are under pressure, and the companies that carried the index for years are no longer giving investors the easy upside they once did.

On the other side, the broader market is still functioning well. Small caps are participating. Regional banks are improving. Industrials, utilities, and infrastructure-linked names are benefiting from real money flows. That is not what a collapsing market looks like.

This is the core paradox investors need to understand right now.

The market is not breaking. Capital is rotating.

Below I’ll break down what is actually happening beneath the surface: why the AI trade has changed form, where the new bottlenecks are, how private and public market valuations are distorting capital allocation, why the $39.3 trillion U.S. debt load matters, and why market breadth still argues for resilience.

Quick summary: what is happening right now

  • The Magnificent Seven are no longer acting like clean AI beneficiaries. They are increasingly being treated as the companies paying for the AI buildout.
  • Hyperscalers are expected to spend roughly $800 billion in capital expenditures this year, with estimates rising to about $1.6 trillion next year.
  • That spending is pushing capital toward industrials, utilities, and power-grid-related businesses rather than toward mega-cap software and platform names.
  • The equal-weight S&P 500 remains strong even as large-cap tech weakens, which signals healthy market breadth.
  • The key bottleneck in AI is not only processors. Memory, especially advanced high-bandwidth memory, has become a major supply choke point.
  • Memory producers with scarce capacity now have substantial pricing power, and that pressure is flowing through to end customers.
  • Public and private markets are both showing valuation distortions, with inflated equity increasingly used as acquisition currency and venture funding concentrating into a narrow set of AI bets.
  • The U.S. debt burden is becoming harder to carry as higher refinancing costs push annual interest expense above $1 trillion.
  • The market’s optimism depends heavily on one macro assumption: that AI and manufacturing investment will materially improve productivity and grow GDP faster than debt.
  • Small caps, healthcare, and regional banks are helping support the broader market even while tech leadership weakens.

Why the AI trade has changed

For the last few years, the standard playbook was simple: buy the biggest technology companies and let index concentration do the work.

That trade has changed.

The most important shift is this: the largest tech companies are no longer being valued primarily as AI winners. They are being evaluated as AI spenders.

That distinction matters.

Companies like Microsoft, Amazon, and Meta operate data-center infrastructure at enormous scale. These hyperscalers are now committing staggering amounts of capital to physical buildout. The spending is not theoretical. It is concrete, steel, copper, power equipment, cooling systems, and processors.

Current estimates put that capex near $800 billion this year and roughly $1.6 trillion next year.

From a financial analysis perspective, that changes how the market values these businesses. Investors like scalable earnings. They like models where revenue rises faster than costs. They like operating leverage.

Right now, hyperscalers are in the opposite phase.

They are in an infrastructure cycle. Upfront costs are surging, while the return profile remains uncertain and delayed. Wall Street is far less forgiving when cash is leaving the balance sheet at that scale without immediate earnings expansion.

That is why parts of big tech have become a drag on the index rather than its engine.

Operating leverage matters more than the AI narrative

This is where many investors get the story wrong.

AI enthusiasm alone does not justify valuation. What matters is whether spending converts into durable profit growth.

Software businesses usually benefit from strong operating leverage because serving one more customer often costs very little. Infrastructure-heavy buildouts do not work that way. When companies are building physical assets at scale, the cost burden arrives immediately.

That is the issue now facing the largest AI spenders.

The market is asking a simple question: how long will it take for these investments to earn an acceptable return?

Until that answer becomes more convincing, investors will continue reallocating toward the businesses getting paid today.

The money is moving to the builders

If hyperscalers are funding the AI boom, then someone else is collecting the cash.

That someone is increasingly found in industrials, utilities, and grid-related infrastructure.

This is the practical side of the AI buildout. Data centers need land development, heavy machinery, electrical systems, transmission capacity, and ongoing maintenance. The companies supplying those inputs are being rewarded because their revenue visibility is improving now, not years from now.

Backlogs for companies tied to this buildout have reportedly grown at rates around 35% to 40% year over year. That is exactly what investors want to see in a capital rotation: real demand, visible pipelines, and tangible order books.

The clean analogy is a gold rush. Early on, investors chase the prospectors. Later, the more reliable returns often come from the businesses selling the tools, transport, and infrastructure.

That is where part of the market has moved.

Why the equal-weight S&P 500 tells a better story than the headline index

Many investors look at the S&P 500 and assume the market’s health is captured by the headline number alone. That is incomplete.

A market-cap-weighted index gives enormous influence to the largest companies. If Apple, Microsoft, Nvidia, Amazon, and a handful of others are weak, the index can look fragile even while many underlying stocks are doing well.

An equal-weight index removes that distortion by treating each constituent the same.

That matters now because the equal-weight S&P 500 has remained near highs even as major tech names weaken. The message is clear: the broader market is healthier than the top-level index suggests.

This is one of the strongest pieces of evidence supporting the rotation thesis. The money is not leaving equities altogether. It is moving away from concentrated mega-cap exposure and toward a wider range of sectors.

The real AI bottleneck is memory

Processors get the headlines. Memory may be the bigger constraint.

AI systems cannot function efficiently without fast access to enormous datasets. That requires specialized memory architecture. In practical terms, the current AI stack depends heavily on DRAM for active memory and NAND for storage, with advanced forms of high-bandwidth memory playing a critical role in high-performance workloads.

This is not a component that can be ramped overnight.

Manufacturing advanced memory takes time, technical precision, and capacity that only a small number of companies can provide. When supply is limited and demand is urgent, pricing power follows.

That is exactly what the market is seeing.

Why memory producers have unusual pricing power

In most hardware markets, high margins attract competition and eventually pressure prices lower. That is the standard commodity cycle.

Memory has often behaved that way historically.

The current debate is whether this cycle is different.

The bullish case says yes. The argument is that advanced memory used in today’s AI systems is not easily replaceable, not quickly expandable, and not yet subject to a practical workaround. If that is true, then producers with available supply can dictate terms.

The bearish case is more traditional. Eventually, software may become more efficient, architecture may evolve, or new supply may enter the market and normalize margins.

That debate remains open, but the present reality is straightforward: supply is constrained, and the companies producing essential memory components are in a strong negotiating position.

One of the most important details is that some of this demand is not being left to open-market uncertainty. Strategic customer agreements are locking in future revenue streams years in advance. That gives producers greater visibility and strengthens pricing discipline.

In other words, this is not just a speculative spike. Parts of the economics are being contractually embedded into the system.

Why this chip squeeze affects more than semiconductor stocks

When a critical component becomes scarce and expensive, the impact spreads far beyond the supplier.

That pressure works through the entire chain:

  • Cloud providers face higher infrastructure costs
  • Hardware makers absorb margin pressure or raise prices
  • Software developers face more expensive deployment economics
  • Consumers eventually pay more for devices and services

That dynamic is already visible. Companies exposed to memory-intensive hardware are raising prices to offset component inflation.

This is also why the issue has macro relevance. In markets with heavy exposure to memory leaders, national equity indexes can become disproportionately sensitive to what appears to be a narrow supply chain variable. A bottleneck in one category of semiconductor can ripple into corporate earnings, export flows, and country-level market performance.

How inflated equity is distorting deals in public markets

Another critical theme is valuation distortion.

When a company trades at an extremely rich public-market valuation, its stock becomes acquisition currency. Management can use a small slice of elevated equity to purchase a real operating business. Economically, that can be smart for the acquirer. Systemically, it can distort pricing across the market.

That is because the deal value depends on the assumption that the buyer’s equity deserves its inflated price.

If the market grants one company a massive valuation multiple, that company effectively gains stronger purchasing power than its cash position alone would imply. It can bid aggressively for strategic assets, push acquisition values higher, and reshape expectations across adjacent sectors.

This matters because investors often misread these transactions as proof of underlying fundamental value, when in many cases the mechanics are driven by equity leverage rather than operating cash generation.

Venture capital is not as adventurous as it looks

The private market tells a similar story.

There is a common assumption that venture capital thrives on broad experimentation and high-risk originality. In practice, current AI funding is heavily concentrated and unusually consensus-driven.

A large share of venture dollars is reportedly flowing into only a handful of firms. Those firms are then writing very large checks into founder teams with elite pedigrees, especially former executives and researchers from major technology platforms.

That is not pure risk-taking. It is institutional fear of missing the next breakout winner.

The result is seed-stage companies receiving valuations that would have looked extreme in prior cycles, often before they have a commercial product or a customer base.

From a capital markets perspective, this creates two problems:

  • It compresses expected future returns because too much value is assigned too early
  • It channels capital into a narrow set of similar bets rather than into broader experimentation

That is not ideal for disciplined investors.

The foundational model race may be less durable than the market assumes

The race to build the best foundational AI model is attracting immense capital, but the durability of that advantage is questionable.

At the base layer of technology, technical advantages can be short-lived. Competitors iterate quickly. Performance improvements get copied. What looks like a breakthrough today may be standard within months.

If that pattern holds, then many companies burning billions to win the model war may discover that the economics are worse than expected.

Owning the base layer is attractive in theory. In practice, it can become a capital-intensive fight with shrinking competitive half-lives.

This is why investors should separate technological excitement from investment quality. They are not the same thing.

Why the application layer may be the better long-term opportunity

The contrarian opportunity is not necessarily in the companies building the most advanced model. It may be in the companies applying AI to specific, practical workflows.

This includes software for ordinary industries such as:

  • HVAC service operations
  • Forestry and agricultural yield analysis
  • Restoration and field-service businesses
  • Back-office process automation for small and mid-sized firms

These businesses do not need trillion-dollar outcomes to generate excellent returns. They need strong unit economics, clear customer pain points, and recurring revenue models.

That is where durable value usually gets built.

Application-layer companies can use AI to improve margins, reduce labor friction, and increase workflow efficiency in parts of the economy that have been underserved by software for years. That is far more tangible than burning enormous sums in a race to commoditize foundational models.

For investors, this is a crucial distinction. The most profitable AI opportunities may not come from the loudest names.

The $39.3 trillion macro backdrop cannot be ignored

None of this is happening in isolation.

The U.S. government is approaching roughly $39.3 trillion in debt, and the deficit trajectory remains severe. The bigger issue is not just the debt total. It is the cost of financing it.

As low-rate debt matures, refinancing occurs at much higher yields. That pushes annual interest expense above $1 trillion, making debt service one of the government’s largest expenditures.

This changes the macro context for every asset class.

Higher debt service reduces fiscal flexibility. It raises the stakes around growth. It increases sensitivity to rates. And it limits the number of politically viable policy paths available.

The government’s options are narrow

At a high level, there are only a few ways to deal with a debt burden of this scale.

  1. Balance the budget and reduce debt directly.

    This would require higher taxes and meaningful cuts to large, politically sensitive spending programs. In practical terms, it is extremely difficult to execute.

  2. Monetize the problem.

    Expanding the money supply to ease the debt burden risks renewed inflation and currency debasement.

  3. Default.

    This would be catastrophic because U.S. Treasuries sit at the core of the global financial system.

There is also a fourth path, even if it is not usually framed this way in public discussion.

The strategy is to outgrow the debt.

The real macro bet: grow GDP faster than debt grows

This is the most important macro link in the current market.

The U.S. does not need to eliminate the debt in nominal terms. It needs the debt burden to become smaller relative to the size of the economy. That means driving GDP growth fast enough that the debt-to-GDP ratio becomes more manageable over time.

This is where the AI and manufacturing buildout takes on national importance.

There is an implicit macro hope that massive private-sector investment in data centers, industrial capacity, and automation will drive a meaningful productivity surge. If that happens, the economy can absorb more debt.

If it does not happen, the policy options become much uglier.

That is why this cycle matters so much. The stakes are larger than technology-sector earnings. This is tied to fiscal sustainability.

Why holding only cash is risky in this environment

If the long-term response to debt pressure involves even partial monetary debasement, then idle cash becomes a weak strategic position.

Cash protects nominal value in the short run. It does not protect purchasing power if inflation accelerates again or if the currency weakens over time.

That is why ownership matters.

Real assets, productive businesses, and equities with pricing power offer a better defense against currency erosion than uninvested cash balances. The exact allocation depends on risk tolerance, but the principle is straightforward: in a debt-heavy, inflation-sensitive regime, preserving purchasing power requires owning assets that can reprice.

So why has the market not collapsed?

Because the market is broader than mega-cap tech.

This is the conclusion many investors miss when they focus too narrowly on a few index leaders. Weakness at the top does not automatically mean weakness everywhere else.

Recent market behavior shows a clear rotation into defensive and cyclical sectors, along with increased participation from smaller companies. That is a healthier setup than a market propped up by only seven stocks.

Even on days when the S&P 500 declines because large technology names are dragging the cap-weighted index lower, a large number of individual stocks can still advance. That is a strong breadth signal.

Breadth does not eliminate risk. It does show that capital is still engaged.

What small caps and healthcare are saying

Two areas help reinforce the point.

First, small-cap participation matters because it reflects improving confidence beyond the market’s largest names. If the Russell 2000 is performing well, investors are taking exposure in more economically sensitive areas that often depend on domestic demand and financing conditions.

Second, healthcare leadership matters because it shows the market is willing to reward earnings durability and sector-specific growth stories outside the AI narrative.

Together, these trends support the idea that this is a rotation market, not a broad liquidation event.

Why regional banks are one of the best signals in the market

If there is one indicator that cuts through the noise, it is the performance of regional and community banks.

These banks are tightly linked to the real economy. They are not financing trillion-dollar model races. They are lending to households, small businesses, local developers, and everyday commercial activity.

Their health depends on things like:

  • Consumers making credit-card payments
  • Auto loans remaining current
  • Small businesses expanding
  • Communities continuing to borrow, spend, and invest

That is why strength in regional banks is so important. It suggests Main Street is still functioning.

Even in a higher-rate environment, even with inflation pressure and slower housing activity, the local economy has remained more resilient than many expected. That resilience helps offset the volatility coming from the top of the technology stack.

What this means for investors

The key takeaway is simple.

Do not confuse index concentration with total market truth.

Right now, the AI trade is not dead. It has changed form. The winners are no longer only the giant platforms and chip headline names. They increasingly include the industrial suppliers, electrical infrastructure firms, utilities, and niche software operators monetizing practical use cases.

At the same time, the macro backdrop remains serious. The debt load is large, interest expense is rising, and the market is effectively betting that AI-led productivity growth will justify this spending cycle.

That makes selectivity critical.

The strongest opportunities are likely to be found where capital spending is translating into real backlog, real pricing power, or real productivity gains. The weakest setups are likely to be those where valuation still assumes perfection while cash burn remains extreme and returns are distant.

The final paradox investors should keep in mind

There is one more important possibility.

If everyday AI applications become efficient enough, businesses may discover they do not need the most expensive frontier infrastructure for many real-world tasks. If that happens, the application layer could become the very force that undermines the economics of the most capital-intensive part of the stack.

That is the paradox.

The success of practical AI could eventually weaken the case for some of the most expensive AI infrastructure spending.

That possibility is not the base case yet, but it belongs on every serious investor’s radar.

Conclusion

The headline message is not that the market is fine and all risks have disappeared. That would be sloppy analysis.

The real message is that the market is more complex and more resilient than a quick glance at mega-cap tech would suggest.

Big tech is under pressure because it is funding a historic infrastructure cycle. Memory has become a strategic choke point with real pricing consequences. Public and private valuations are distorting capital allocation. The U.S. debt burden is forcing a national wager on growth. And yet, beneath all of that, the broader market is still participating.

That is the Great Rotation.

The top of the market may look unwell. The rest of the system is still running.

Sources