Steve Eisman made his name betting against subprime mortgages before the 2008 financial crisis, a trade chronicled in the book and film “The Big Short.” On Aug. 13, he turned his attention to the AI boom, and his warning was specific: the market’s enthusiasm now depends on the fortunes of two private companies, OpenAI and Anthropic.
Speaking at an investor conference, Eisman said a large portion of the AI revenue reported by the biggest technology companies flows through those two firms. If either stumbles, he argued, the ripple effects would hit the revenue lines of companies whose AI businesses have become central to their valuations. The concentration risk, he said, is the thing the market is not pricing.
The argument rests on how the AI economy is actually structured. The hyperscalers and enterprise software giants that report AI revenue largely sell access to models built by OpenAI and Anthropic, or resell those models through their cloud platforms. When a customer spends on AI, much of that spending ultimately traces back to the two labs. Eisman’s point is that the foundation of the AI trade sits in companies that are not public, do not report detailed finances, and cannot be held accountable by shareholders.
Eisman, who manages money at Neuberger Berman, was careful to separate his skepticism about the trade from skepticism about the technology. He said AI will change the world, a concession that matters because it shows how much the debate has moved. The argument among serious investors is no longer whether AI matters; it is when the market’s expectations will meet reality, and at what valuation.
He also pointed to a risk that gets less attention in American boardrooms: Chinese AI models. Eisman called the Chinese competition the Achilles’ heel of the trade, arguing that open-source Chinese models, built with fewer resources and distributed freely, could compress the pricing power of American labs faster than expected. If frontier models become cheap or free, the revenue concentration he identified at OpenAI and Anthropic becomes a liability rather than a moat.
The market reaction to his remarks was notable mainly for its absence. AI stocks barely moved on the speech, a sign that a bear case from a famous investor no longer surprises the market the way it once did. The sector has absorbed warnings before, from economists, academics and short sellers, and each time prices have resumed their climb. Eisman’s credibility gives his comments weight, but the market has heard versions of this argument for two years.
The concentration point, however, is harder to dismiss. OpenAI’s revenue has grown at a pace that makes it one of the largest software companies in the world by sales, and Anthropic has scaled similarly, both while remaining private. The public companies that report AI-driven growth are, in large part, distribution channels for the two labs. The enterprise software vendors, cloud providers and device makers all earn money from AI, but the models at the center of the ecosystem belong to two firms.
That structure creates a specific kind of risk. A product failure, a leadership crisis or a regulatory action at either lab would hit the revenue of multiple public companies simultaneously, Eisman argued, and investors have no direct way to hedge that exposure because the labs are not listed. The public market is, in effect, long two private companies it cannot trade.
Eisman’s second point, about Chinese competition, draws on a track record of identifying risks that others dismissed. He was early and right on subprime, and later on European banks, and his willingness to take unpopular positions gives his AI comments a hearing they might not otherwise get. Whether he is right this time depends on whether Chinese models actually close the gap with American ones, a question the industry itself cannot answer with confidence.
The most useful part of his speech may be its framing of the debate. When a famous bear says AI will change the world, the disagreement among investors narrows to two questions: timing and valuation. Bulls believe the revenue will arrive quickly enough to justify today’s prices; bears believe the gap will be closed by Chinese competitors or by the simple reality that expectations have run ahead of adoption. Both sides can be right about the technology and wrong about the money.
The conference appearance also gave Eisman a chance to distinguish his current view from his famous short. He has said repeatedly that he is not shorting AI stocks, because the trade he identified in 2008, a structural flaw in securities that had to default, does not exist here. The AI boom is built on real revenue and real products, he acknowledged, which makes it a valuation debate rather than a fraud. That distinction is worth noting: the man who profited from the last great financial collapse says this is a different animal entirely.
Eisman’s warning lands at a moment when the AI trade is still climbing. He did not call a crash, and he did not say the technology is a bubble in the sense that it will prove worthless. His claim is narrower and harder to refute: that the market has concentrated enormous value in two private companies, and that concentration is the risk nobody can hedge. Investors who heard him walked out with a question to answer rather than a position to copy.
