Steve Eisman of “The Big Short” fame said he is starting to worry that the artificial intelligence boom may rest on shakier ground than investors assume. “The Real Eisman Playbook” podcast host and former Neuberger Berman senior portfolio manager said he encountered a theory suggesting that as large language models continue to scale, their performance gains could begin to diminish, challenging the premise that ever-larger models will keep delivering breakthrough improvements. “The large language models, as they keep scaling, which is the model that everybody has, will start to lose their efficaciousness,” Eisman said Thursday on CNBC’s ” Squawk Box. ” “The improvement is gonna slow as opposed to increase … At some point, companies like Microsoft — if this becomes true — they’re going to start buying fewer chips.” This idea runs counter to one of Wall Street’s most widely accepted foundations of the AI trade — that rising model complexity will justify the enormous spending on computing power that’s driving chip demand. “Now, I’m not here to panic people, but if this argument is right, you’re going to start to see more and more people’s start to say this,” he said. “This is the foundational argument that everybody should be focused on all the time.” Eisman said he’s staying put right now, still owning most of the key AI players like Nvidia , Microsoft, and Meta Platforms . Eisman framed the concern as a foundational risk, comparing it to the flawed assumption underpinning the pre-crisis mortgage market circa 2006-2008. “It’s like the foundational argument before the great financial crisis, where I’ve eventually figured out that the entire mortgage fixed income market rested on one assumption, which was housing prices can’t go down,” he said. “And once that assumption got pulled out, the whole edifice collapsed.” Michael Burry, also of the “Big Short” fame, has for months been warning that AI demand could be overstated and that the economics of the current spending surge. He is taking his angst even further and betting against AI.
