Is the AI Capex Boom Starting to Crack?
Oracle just had its worst week since the depths of the dot-com bust. Apple is hiking prices on its products because memory chips are suddenly expensive. A prominent venture capitalist went on CNBC and said the "all eggs in AI basket" trade is breaking. And Qualcomm promised $15 billion in future AI chip sales—with the small caveat that the chips don't exist yet.
This isn't a random collection of tech headlines. It's a single, uncomfortable question for every investor who has ridden the AI wave: What happens when the cost of building the future starts to outweigh the returns from selling it?
The AI narrative has been remarkably resilient. For two years, the market has rewarded any company that could credibly claim to be building AI infrastructure. Capital was cheap, demand seemed infinite, and the future was a discounting machine. But the week ending June 26, 2026, felt different. The discount rate is no longer your friend.
Consider Oracle. Its stock fell 19% in a single week—a drop not seen since August 2001. The trigger? Capital expenditures surged 162% in the latest fiscal year, producing nearly $24 billion in negative free cash flow against $130 billion in debt. Oracle is borrowing heavily to build data centers. The market looked at the bill and asked: *When does this become a cash-flow business, not a construction project?*
That same tension is rippling through the entire ecosystem. Amazon, through AWS, just raised key AI cloud prices by 20%—on top of a 15% increase in January. The official reason: soaring memory chip costs and strong AI demand. But read between the lines. Amazon is passing through its own cost inflation to customers. The AI buildout is becoming an expense burden that gets handed downstream. For every company renting compute from AWS, the math just got harder.
Apple, meanwhile, is feeling the squeeze from the same memory chip inflation. The company that prides itself on supply chain mastery is now hiking consumer prices. When Apple—the ultimate pricing power story—has to pass along component costs, it signals that the AI capex boom is creating real, tangible inflation in the hardware layer. The question isn't whether AI is real; it's whether the cost of the raw materials to run it is being fully priced into the stocks that depend on it.
Then there's Qualcomm. At its Investor Day, CEO Cristiano Amon promised $15 billion in AI chip sales by 2029. The catch, as one analyst noted, is that the chips don't exist yet. This is the purest expression of the AI capex boom's forward-looking optimism. But in a world where the 10-year Treasury yield is at 4.38% and the Fed is still talking about rate cuts as a future possibility, promises that far out get discounted heavily. The market punished Qualcomm with a 15% weekly drop.
What's striking is not that any single company is in trouble. It's that the entire thesis is being tested simultaneously. The AI winners—the companies with actual cash flows and pricing power—are starting to look different from the AI spenders—the debt-dependent builders and demand-dependent chip suppliers.
Sarah Kunst of Cleo Capital captured the mood when she argued that the "all eggs in AI basket" trade is cracking. She's not saying AI is a bubble. She's saying that the *financing* of AI—the assumption that infinite capital should flow to infinite buildout—is hitting a wall. When capital costs rise, the companies that can fund their own growth without borrowing have a structural advantage. The companies that need to issue debt or promise future revenue to keep building face a much higher bar.
This is where the market may be mispricing something. The AI boom has been treated as a monolith. But the week's action suggests a divergence is forming. The cash-flow funders—companies like Microsoft and Meta, which generate enormous free cash flow and can self-fund their AI ambitions—may weather the higher-rate environment better than the debt-dependent builders. Oracle's plunge is a warning shot, not a death knell.
The real investing question isn't whether AI will change the world. It almost certainly will. The question is: *In a world of expensive money, who captures the value, and who just pays the bills?*
The market is starting to ask that question out loud. The answer will determine whether the next leg of this cycle belongs to the builders or the owners of the infrastructure they're building on top of.