AI Stocks Stumble as Investors Weigh the Cost of the Data Center Boom

The companies at the center of the artificial intelligence boom had a rough week on Wall Street, as investors grew more anxious about how much borrowed money is going into data centers and chips, and whether stock prices that soared during a years-long frenzy have run too far ahead of profits.

The sharpest drop came Thursday. The S&P 500 fell 1.2 percent and the technology-heavy Nasdaq composite tumbled 2.6 percent, according to market reports from NPR and Euronews. The losses were concentrated in tech: most stocks in the S&P 500 actually rose, and the Dow Jones Industrial Average finished essentially flat, up less than 0.1 percent.

Strong results were not enough to protect shares. Lam Research, which makes equipment used to produce semiconductors, reported profit and revenue slightly better than analysts expected, yet its stock fell almost 12 percent, according to The Guardian. Meta Platforms lost 7 percent after reporting stronger-than-expected revenue but profit that fell short; the company also said it would introduce a new class of nonvoting shares. Nvidia, the most valuable company on Wall Street, fell 4.4 percent and was the biggest single drag on the S&P 500.

The most closely watched trouble spot has been CoreWeave, which borrows heavily to buy Nvidia chips that it rents to customers through the cloud. Its shares fell about 10 percent on Thursday amid worries about its debt, and NPR’s Planet Money flagged a distressing credit warning for the company among the week’s notable economic indicators.

The selloff began in Asia, where markets are heavy with chip and electronics companies. South Korea’s Kospi fell 3.5 percent and Japan’s Nikkei 225 2.7 percent on Thursday, The Guardian reported. Rising oil prices added to the pressure: Brent crude jumped about 4 percent that day to its highest level in five months on fears about the war with Iran, and on Friday it climbed further, passing $117 a barrel at one point, The Guardian reported.

Borrowing costs are a central part of the story. Worries about large government debts, persistent inflation and a still-solid American economy have pushed bond yields to their highest levels in decades, which makes the debt that funds data center construction more expensive. The yield on the 10-year Treasury had topped 5.3 percent earlier this month. On Thursday, though, it fell sharply to about 4.97 percent as investors moved into government bonds as a safer place for their money.

Not everyone sees a bubble. The industry’s defenders point out that AI companies continue to report growth, and a strong economy has lifted corporate profits broadly; The Guardian noted that the stock market set a record high earlier this month despite the many worries around it.

For now, the market is sending a mixed signal: investors are still rewarding the broad economy while demanding more proof that the enormous sums flowing into AI infrastructure will earn a return.