Steve Eisman, of “The Big Short” fame, said he was starting to worry that the artificial intelligence boom was on steadier footing than investors assumed. The host of the podcast “The Real Aisman Playbook” and former senior portfolio manager at Neuberger Berman said he has come across theories that suggest performance gains may start to diminish as large language models continue to scale, calling into question the assumption that ever-larger models will continue to provide breakthrough improvements. “As large-scale language models continue to expand, the models that everyone has, they’re going to start to lose their effectiveness,” Eisman said on CNBC’s “Squawk Box” Thursday. “I think the improvements will slow down, not increase… At some point, companies like Microsoft — if this is true — they’ll start buying fewer chips.” This idea runs counter to one of the most widely accepted cornerstones of AI trading on Wall Street: the idea that increased model complexity justifies massive spending on computing power to drive demand for chips. “I’m not here to panic people, but if this argument is correct, you’ll see more and more people start saying this,” he said. “This is a fundamental discussion that everyone should always be paying attention to,” said Eisman, who is now in office and still owns most of the major AI players, including Nvidia, Microsoft and Meta Platforms. Eisman framed this concern as a fundamental risk, comparing it to the flawed assumptions that underpinned the mortgage market before the financial crisis around 2006-2008. “This was kind of the fundamental argument before the Great Financial Crisis, and what we ended up finding out was that the entire mortgage bond market was based on one assumption: house prices weren’t going to go down,” he said. “And when that assumption went wrong, the whole building collapsed.” Michael Barry, also of “The Big Short” fame, has been warning for months that demand for AI may be overstated and that current spending is exploding economically. He deepens his anxiety and bets against the AI.
