AI shares rebound after selloff erased more than $2 trillion as data-center votes unsettle investors

PIMCO’s Cantrill says an AI wobble could become a headwind for growth

AI shares rebounded Friday after a one-day selloff wiped out more than $2 trillion in value, but investors remain shaken, NPR senior economics and finance editor Rafael Nam said.

Nam said the scale of the losses was “pretty staggering” — the kind “you only see every once in a while” — and that traders had already felt uneasy about the AI trade all year long, with a lot of money riding on the technology and concerns about whether the spending will pay off. It is not clear whether the market is in the middle of a tech bubble that is about to burst, NPR’s report noted.

The midterm results sit at the center of the anxiety. Nam said data centers are critical to AI companies, and that when a wave of Democrats is elected talking about creating moratoriums or slowing down the development of data centers, “investors can freak out.”

NPR tech correspondent John Ruwitch reported a construction boom behind the fears: trillions of dollars committed to new AI data centers, and local nervousness about water and electricity costs and about jobs.

Ruwitch said he attended a town hall the night before in San Jose, California, in the heart of Silicon Valley, where plans call for as many as 13 new data centers. Residents like Justin Brown say things are moving way too fast, he reported. “Like, let’s actually have a debate, and let’s talk about this and see if there’s some sort of common ground so the residents can get something out of this,” Brown said.

Data centers were on the ballot nationwide on Tuesday, Ruwitch reported. Candidates elected in places like Texas, Ohio and South Carolina have spoken out against them, and dozens of local measures across 13 states blocked or restricted future data center development.

The outcome suggests it will be harder going forward for AI companies to build data centers, Ruwitch reported. The companies have invested what he called eye-watering amounts of money in the computing power to train new AI models — data centers full of high-end chips. There is already a shortage of that computing power, and if leading AI companies do not build more data centers, their plans to build better AI and their business plans “get gummed up.”

Nam said that if AI companies cannot develop AI models as fast as they want to, they may not see the profits they are promising markets — leaving investors asking how soon they will make their money.

So much of the stock market now rides on the AI trade that it could become a bigger deal than market losses alone, Ruwitch said. Libby Cantrill, head of public policy at the investment management company PIMCO, said “The U.S. economy is an AI economy.” If the AI story starts to wobble, she said, it could be a headwind for U.S. economic growth.

Just about everybody believes there is bound to be some pain in developing AI, as with the dot-com bust at the start of the internet, Nam said. The deeper worry is the industry’s structure: AI companies are not only raising massive amounts of debt, many are also investing in each other and doing business with each other. Nvidia, the big chip company, has invested in startups like OpenAI and Anthropic, and those startups in turn buy massive amounts of chips from Nvidia. The circular arrangement works fine when things are going well, Nam said, but if one company gets into serious trouble, it could set off a chain reaction that investors fear could spark a global financial crisis like the one in 2007.

Ruwitch pointed to what he called a contradiction baked into the promise of AI: the technology is powerful, but no one yet knows the best ways to use it. The companies are placing huge bets that some of the best use cases will be things that cannot be imagined today. It might take years for breakthrough applications to appear and to make money, he said. It might never happen — and the possibility of a data center slowdown complicates that picture.

What happens next is unknown. “We just don’t know what will happen,” Nam said. “Voters have spoken, but we need to see what elected officials will actually do.”

One big thing to watch, Nam said, is the upcoming initial public offering by Anthropic, expected this month. The price those shares sell at on their first day and after could say a lot about investor appetite for companies in AI, Ruwitch said. He noted that Anthropic is a financial supporter of NPR.

Nam ended on the note many experts offer: the economy and the markets may have become too dependent on AI. A pullback in data centers and in markets may be painful, he said, but it could ultimately help rebalance things. The dot-com bust, he noted, cleared out weak companies like pets.com and the financial froth — and today, he said, you cannot imagine a life without the internet.