Eisman says AI boom has Achilles' heel, but no short yet
Steve Eisman says AI depends heavily on OpenAI and Anthropic and calls that an “Achilles’ heel,” but says it is too early to short the boom without evidence of deterioration.
“Big Short” investor Steve Eisman says the AI boom has an “Achilles’ heel”: its enormous dependence on OpenAI and Anthropic.
But he isn’t ready to short the AI boom yet, because investors lack what he had before betting against subprime mortgages: evidence the breakdown has actually begun.
Eisman Says AI Bears Don’t Have the Data Yet Speaking on Friday’s episode of his Weekly Wrap podcast, Eisman recalled that before shorting subprime, his team bought access to Moody’s securitization database.
Month after month, the delinquency data deteriorated and confirmed his thesis. “There is no such data set with respect to AI,” he said.
Both labs are private, leaving investors to infer their economics from funding rounds and spending commitments.
That is why Eisman considers calls for a crash “premature.” He remains, in his words, “quite long.” OpenAI and Anthropic Are the ‘Achilles’ Heel’ Still, the investor sees a real bear case.
There “don’t seem to be any moats around LLMs,” he said, as users switch between models constantly and Chinese open-weight rivals are far cheaper.
That combination, he warned, could eventually trigger a price war.
The stakes are enormous.
Research estimates cited by Eisman suggest OpenAI and Anthropic account for roughly 70% of AI revenue across Microsoft Corp. (NASDAQ: MSFT ), Amazon.com Inc. (NASDAQ: AMZN ) and Alphabet Inc. (NASDAQ: GOOGL ), and 25% to 35% of their cloud revenue.
Eisman also said roughly half of the approximately $600 billion backlog at Oracle Corp. (NYSE: ORCL ) comes from OpenAI.
Eisman called that dependency “huge and quite scary.” If a price war undermines the labs’ economics, hyperscalers may pull back on capex, and “the entire AI chain goes into reverse.” The IPOs Could Put Eisman’s Theory to the Test “When OpenAI and Anthropic go public, we will have some real data,” Eisman said.
Both Anthropic and OpenAI have already filed confidentially for U.S.
IPOs, but prediction-market traders expect Anthropic to reach public markets first.
Polymarket gives Anthropic a 90% chance of beating OpenAI to an IPO.
In a separate market, traders price a 70% chance that Anthropic completes its IPO before Nov.
1, rising to 85% before Dec.
31.
OpenAI looks further away.
The New York Times reported in June that the company was leaning toward delaying its IPO until 2027, and on Polymarket, there’s just a 20% chance of an OpenAI IPO this year.
Until Then, the Boom Keeps Running So far, Eisman sees little evidence that the breaking point has arrived.
Hyperscalers are still spending heavily, while Nvidia (NASDAQ: NVDA ) is working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilize more than $500 billion for AI infrastructure.
Anthropic is also still growing rapidly, projecting $190 billion to $200 billion of 2028 revenue, up from a $47 billion annualized run rate in May, That is why Eisman isn’t shorting yet.
Until trouble at OpenAI and Anthropic begins to “metastasize,” he said, the AI story can keep running.
Image: Shutterstock Read Also: Elon Musk Says AI Could Hit Earth's Power Wall in 2029: 'Orbital Compute' Will Be the Only Way to Scale