AI Debt Strains Tech Stocks as Washington Backs Buildout

The Nasdaq Composite fell roughly 2.5% on Friday, while the S&P 500 dropped 1.4%, as investors assessed the cost of an AI infrastructure boom Washington has made a national priority. Both indexes had hit record highs 10 days earlier. Now, higher interest rates, rising AI-related debt and political disputes over data centers are weighing on technology shares.

Debt Behind the AI Boom

CoreWeave, a neocloud company backed by Nvidia that rents computing power to AI businesses, sank 14% on Friday. It lost one-quarter of its value during the week and stood more than 60% below its peak shortly after its March 2025 initial public offering. The company carries tens of billions of dollars in debt to fund growth, and its bond yields recently climbed to levels that signal distress or greater default risk.

“It's unclear that (CoreWeave is) going to get a sufficiently good return to pay down the debt that they borrowed,” Gil Luria, head of technology research at D.A. Davidson, told CNN. He called neocloud companies “the subprime tranche of the AI trade” as investors question how quickly the buildout is proceeding. Luria said AI demand is real and growing, but also pointed to concerns about unsustainable financing and steep interest rates.

Nvidia CEO Jensen Huang told Bloomberg on Thursday that he was optimistic about CoreWeave, and the company’s shares rallied that day. The gains didn't last. Higher borrowing costs are pressuring debt-dependent parts of the AI buildout, while investors question whether projected demand will generate returns large enough to justify the financing.

A National Priority, With a Rising Price

A New York Times DealBook item said the largest technology companies would spend around $800 billion on AI infrastructure in 2026. Consensus forecasts put spending above $1 trillion for 2027. Goldman Sachs said that would account for a larger share of gross domestic product than any investment cycle since the railroads.

The DealBook item said Washington had made AI a national priority, while asking how much the buildout would cost and whether companies could do it for less. It also noted that companies were using tax credits to help fund spending. The source reports no demographic effects or transfer of national authority associated with the spending; the documented policy question concerns the scale of investment and who bears its costs.

Nvidia reported a gross margin of 75% in the previous quarter, while Micron, which sells memory chips, reported a margin of nearly 87%. Micron lost money on memory chips three years earlier. DealBook said shortages set prices and high margins typically attract competitors, but competition hasn't yet cut margins in the AI buildout, where the signal has been visible since 2023.

Chipmakers Face New Risks

A closely watched semiconductor-stock index dropped 3.9% Friday, its eighth straight daily decline. Investors had driven chip stocks higher this year on expectations of a major AI buildout and shortages that could deliver substantial profits. Now they're weighing higher costs, political disputes over data centers and the possibility that cheaper, efficient Chinese chips could erode U.S. and South Korean companies’ market share and pricing power.

Chinese memory-chip maker CXMT announced earlier in the week that it had made progress on advanced memory chips. Investors had expected U.S. and South Korean chipmakers to keep their dominant market share. South Korea's benchmark Kospi, heavily weighted toward SK Hynix and Samsung Electronics, fell 14.5% during the week, its worst weekly performance since 2008.

The semiconductor index had fallen more than 20% from its June peak but remained up almost 60% year to date. Semiconductor companies made up roughly one-fifth of the S&P 500's market value. The VIX, Wall Street's fear gauge, jumped 30% Friday to above 20 points, signaling elevated volatility.

“The market sells first, asks questions later,” said Thomas Martin, portfolio manager at Globalt Investments. The DealBook item cited AMD, custom silicon from Amazon and Chinese open-weight models as developments that could eventually cut AI costs, though that would take time. It also said Federal Reserve chairman Kevin Warsh suggested he wouldn't push to raise borrowing rates that month, while persistent inflation could force the Fed's hand.