Steve Eisman, famous for betting against the housing market before the 2008 crash, sees a 70% chance the AI boom ends in a bear market or worse.
He sees two ways it could go wrong: AI succeeds while investors lose money, or the technology fails to live up to its hype.
In the latest episode of The Real Eisman Playbook, he assigns 40% odds to a bear market caused by excessive investment despite AI's success. Another 30% goes to a scenario where AI proves overhyped, triggering a market crash and U.S. recession.
30% goes to the bull case.
Eisman compares AI to the railroads and early internet, technologies that transformed the world while leaving a trail of bankruptcies.
Why AI Could Succeed and Stocks Still Crash
Eisman cited expectations that technology giants will spend roughly $850 billion on AI infrastructure this year. He also pointed to research suggesting US data center capacity could reach twice actual demand by late 2027.
"There is simply way too much capex chasing demand," he said of his 40% scenario.
Competition adds another threat. Eisman argues OpenAI and Anthropic have little protection against cheaper Chinese models, leaving customers free to switch providers.
Their models are superior, he conceded, but "you don't need a Mercedes-Benz for every trip."
What Happens if the Money Stops?
Oracle Corp. (NYSE:ORCL) reported $664 billion in contracted backlog in September. Roughly half is tied to OpenAI, Eisman said, potentially exposing Oracle if the developer struggles.
Banks are also growing more cautious about financing AI infrastructure, Reuters reported Monday.
OpenAI's June-quarter revenue grew 18% from the March quarter to $6.7 billion, while its operating loss widened by $3 billion to $12.3 billion. Anthropic's revenue more than doubled to $11.6 billion over the same period, The Wall Street Journal reported.
"I'm starting to think that the demise of OpenAI would push the U.S. into an immediate recession," Eisman said.
He stressed, however, that he isn't predicting OpenAI will fail.
What Prediction Markets Say
Traders on Polymarket gave an AI bubble burst roughly 6% odds by December 2026 and around 23% by June 2027 on Tuesday. The contract requires three severe events within 90 days, such as major chip-stock declines or an AI developer's bankruptcy, so a milder bear market like Eisman's 40% scenario might not trigger it.
Traders also appear to share Eisman's doubts about OpenAI. Polymarket gives Alphabet (NASDAQ:GOOGL) roughly 71% odds of having the top-ranked AI model at year-end, compared with 21% for Anthropic and just 2.5% for OpenAI.
Eisman sees greater appeal in the companies supplying AI's "picks and shovels," including Nvidia (NASDAQ:NVDA) and Micron Technology (NASDAQ:MU).
"These businesses don't necessarily need to predict which LLM ultimately wins," he said.