The AI Capex Question That Could Unravel the Market Rally

The Cracks Are Real

One week after the S&P 500 touched fresh all-time highs, the market is already giving back ground. The index sat roughly 1% below that peak as of Thursday’s close, while the Nasdaq had surrendered more than 3%. This is not a panic. It is something more interesting — and potentially more dangerous for investors who have grown accustomed to AI being the one narrative strong enough to power rallies through rough macro conditions.

What is driving the sell-off is a convergence of signals that, taken individually, might be dismissed as noise. Taken together, they point to a growing unease on Wall Street about whether the trillion-dollar bet on artificial intelligence is beginning to outpace the revenue it can justify. The first domino fell on Wednesday when the Financial Times reported that quarterly revenue growth at Anthropic had slowed. The second came the next day when economist Ed Yardeni lowered his year-end S&P 500 target from 8,400 to 7,900, citing rising caution around the AI trade and lingering inflation risks. The S&P was trading near 7,600 at that point — already meaningfully below where most bulls thought it would finish the year.

Who Is Most Exposed

If AI capital expenditure starts to bend, the pain will not be distributed evenly across the market. The companies most exposed fall into two distinct groups: the semiconductor suppliers riding the infrastructure buildout, and the cloud giants whose balance sheets are leveraged to data center demand.

NVIDIA remains the canary in this coal mine. The company’s revenue has been sustained by an extraordinary run of data center GPU sales, with customers race to secure supply ahead of anticipated chip shortages. If hyperscalers and AI startups begin to question the return on their infrastructure bets, NVIDIA’s next few quarters will look dramatically different from the last two years. AMD and Broadcom face similar exposure, though on a smaller scale. Any slowdown in capex directly threatens the revenue models these companies have built over the past 18 months.

On the cloud side, Microsoft, Alphabet, Amazon, and Meta have all committed to multi-year, multi-billion-dollar infrastructure programs. The concern is not that these companies will stop spending — they likely won’t. The concern is that the spending will eventually need to generate operating returns that justify the depreciation schedules they are writing. If AI revenue does not keep pace, margins compress across the board.

Perhaps the most under-discussed exposure lives in the bond market. Data center financing has relied heavily on structured debt, and investors are already selling off those instruments over fears that AI revenues will be insufficient to cover the borrowed funds. This is a quiet stress signal that has not yet appeared in headline equity prices but could accelerate if the trend continues.

The China Variable

Complicating the picture is a geopolitical dimension that most US investors are still underweighting. Chinese memory technology companies are moving aggressively into mass production, and the prospect of an oversupply glut is creating anxiety among semiconductor suppliers. The same dynamics that created the current chip shortage could reverse quickly if Beijing’s industrial policy succeeds in flooding the market.

This is not a hypothetical. Memory chip cycles have a history of being brutal, and the transition from shortage to surplus in 2023 showed how fast sentiment can flip. If Chinese producers capture meaningful market share in the next 12 to 18 months, the revenue outlook for Western memory and semiconductor companies deteriorates further, adding another pressure point to an already stretched valuation framework.

The Macro Implication

The reason any of this matters beyond tech portfolios is that nearly half of all U.S. economic growth this year has come from tech-related spending. That is a concentration risk in a market where most Americans do not own tech stocks directly, let alone work in AI. The wealth effect generated by those gains has been a key prop behind steady consumer spending and employment numbers, even as inflation and declining real earnings continue to pressure lower- and middle-income households.

Wells Fargo CEO Charlie Scharf defended the trajectory this week, arguing that AI investment, high employment and consistent wage growth outweigh the risks from higher interest rates and elevated energy prices. But the data from the Bureau of Labor Statistics on Thursday told a different story. Business inflation hit its highest level since December 2022, a reading that typically precedes consumer price pressure as costs flow downstream. Citi analysts flagged the same concern, listing higher rates, elevated energy costs, slow job growth and equity sell-offs as compounding downside risks to consumer spending.

The chain of causation is straightforward and fragile: tech spending drives earnings, earnings drive stocks, stocks drive confidence and consumption, consumption drives growth. Break any link in that chain and the whole model stutters. That is what makes the current pullback worth watching closely — not because it signals an immediate crisis, but because it reveals how narrow the foundation of the market rally has become.

What Comes Next

The most likely scenario is not a collapse but a reckoning. AI spending will continue at a high level through 2025, but the pace of revenue growth required to justify it will face increasing scrutiny from investors who are beginning to demand proof of returns. Companies that cannot demonstrate that their infrastructure investments are translating into monetizable output will face multiple compression. The market has priced those investments as permanent, almost risk-free additions to future cash flows. That pricing is about to be tested.

For the broader market, the lesson is simpler. The S&P 500’s rally has been unusually narrow, sustained almost entirely by a handful of mega-cap tech names and their suppliers. When conviction weakens in that core thesis, there is not a deep bench of sectors ready to absorb the rotation. That is why a 1% or 3% pullback this week could evolve into something more meaningful if the next round of earnings reports fails to reassure.

Ed Yardeni’s target reduction to 7,900 is a reminder that some of the most respected voices on Wall Street are already adjusting their expectations. The question for investors is whether they are adjusting fast enough.