Chips, Capex, and the Fed: Is the AI Trade Unraveling?
The AI trade is suddenly under attack from all sides: falling chips, soaring capex, and a hawkish Fed. This week, semiconductor stocks resumed their slide, with memory favorites like Micron and Sandisk leading the declines. At the same time, Moody's warned that "unprecedented" AI spending threatens the credit quality of hyperscalers like Amazon, Meta, and Alphabet. And bond yields climbed as fixed-income investors demanded more compensation for lending to companies pouring billions into data centers.
The "huh, interesting" moment isn't that any single one of these forces is new. It's that they're converging simultaneously, raising a question that cuts to the core of the bull case: Is the AI trade breaking down, or is the market simply repricing who captures the value?
Let's start with the chips. Micron fell 12% in the past month, and the broader semiconductor weakness isn't tied to a single "smoking gun" catalyst. Instead, it's a stew of concerns: Chinese memory developments, Korean stock-market weakness, and Intel's inability to sustain post-earnings gains. When the foundational layer of the AI stack—the chips that power training and inference—starts to wobble, it suggests demand expectations are cooling, or at least becoming more discriminating.
But here's the tension: the hyperscalers are still spending. Alphabet lifted its capex forecast, and the Magnificent Seven collectively lost $787 billion in a single day as investors recoiled at the scale of the buildout. Moody's highlighted that the AI buildout is eroding free cash flow and increasing balance-sheet risk at companies like Alphabet and Microsoft. The bond market is listening: Google, Amazon, and Meta are seeing credit spreads widen as fixed-income investors demand more reward.
This is where the investing question gets sharp. The market is pricing in two contradictory narratives simultaneously. On one hand, chip stocks suggest AI demand is becoming more selective—maybe even peaking. On the other, hyperscaler capex suggests the buildout is accelerating, with no end in sight. Which one is wrong?
The answer likely lies in the cost of money. The Fed is expected to hold rates at 3.75% next week, but the yield curve is steepening—the 10-year hit 4.69% this week. Expensive money changes everything. It means the "AI winners" are no longer just the companies with the best technology. They're the ones with the balance sheets to fund the buildout without diluting shareholders or loading up on debt.
The distinction that matters now is between "AI funders" and "AI spenders." Funders are companies like Nvidia, which sells the picks and shovels and generates enormous cash flow from the buildout. Nvidia's stock was up 2% this week even as the broader tech sector struggled—a sign that the market still rewards the company that gets paid upfront. Spenders are the hyperscalers—Alphabet, Amazon, Meta, Microsoft—that are pouring capital into data centers with uncertain payback periods. Their stocks are getting hammered because investors are asking: When does this spending translate into returns?
The bond market is asking the same question. When credit spreads widen on companies with pristine balance sheets, it's a warning that fixed-income investors see rising risk. If the cost of capital continues to climb, the spenders may eventually be forced to slow down. That would ripple back to the chip makers, creating a feedback loop that could indeed unravel the AI trade.
But here's what the market may be mispricing: the AI buildout is not a monolith. The hyperscalers are spending on different things—some on training infrastructure, others on inference, still others on enterprise applications. The chips that power training (like the ones Micron makes) may face cyclical headwinds, while chips for inference (like Nvidia's) may see sustained demand as AI moves from training to deployment. The capex that spooks bond markets today may be the foundation for massive free cash flow tomorrow—if the spending is disciplined.
The reframed question for investors isn't "Is AI real?" It's "Who gets paid in a world where money costs 4.5%?" The AI trade may not be unraveling. It may be rotating—from the spenders to the funders, from the dreamers to the disciplined. The market is forcing a distinction between technology promise and financial reality. That's not a crash. That's a correction in expectations.
The next time you see a headline about chip stocks sliding or bond yields rising, ask yourself: Is this the end of the AI trade, or is it the beginning of a more mature, more selective one? The answer determines not just which stocks to own, but how to think about value creation in a higher-rate world.