AI Boom Warning: Steve Eisman Sounds Alarm Over Large Language Models' Scaling Limitations

‘Big short’ Steve Eisman, the investor who famously predicted the 2008 housing crash is sounding a quiet alarm about the foundations of the artificial intelligence boom. On CNBC’s Squawk Box, he said that he came across a new theory that challenges one of Wall Street’s deepest beliefs about AI, that bigger models will always produce dramatically better results. Eisman said the concern is simple but potentially enormous. “The large language models, as they keep scaling, which is the model that everybody has, will start to lose their efficaciousness. The improvement is gonna slow as opposed to increase,” he said on CNBC. “If that happens, the entire economic case for the AI spending boom could weaken. Tech giants like Microsoft, which are pouring billions into chips, data centers, and infrastructure may find that returns do not justify the enormous costs. As Eisman put it,“At some point, companies like Microsoft if this becomes true they’re going to start buying fewer chips.” This would strike at the heart of the AI trade, where investor enthusiasm is built on the assumption that ever-larger AI models will keep delivering breakthrough improvements. Eisman stressed he is not predicting a collapse, but he believes this line of thinking could spread fast if proven true. “I’m not here to panic people, but if this argument is right, you’re going to start to see more and more people start to say this,” he told on CNBC. He argues that markets are focusing too much on whether AI is a bubble rather than asking the more important question, Can all this spending actually produce measurable returns? “I think what people are focusing too much on is the whole debate that’s going on is actually much more foundational,” he added. For Eisman, the real issue is figuring out how AI will deliver profits, not just whether the technology is exciting. Eisman is particularly sensitive to “foundational assumptions.” The last time Wall Street clung to one, it ended in disaster. “The entire mortgage fixed income market rested on one assumption, which was housing prices can’t go down,” he said on CNBC. “And once that assumption got pulled out, the whole edifice collapsed.” He worries AI could have a similar blind spot, the assumption that LLMs will keep scaling forever and keep delivering steep performance gains. Eisman referenced the work of NYU professor Gary Marcus, a long-time critic of current AI approaches. Marcus has repeatedly argued that LLMs may be approaching a performance ceiling, that simply scaling up size and data is delivering diminishing returns. Eisman seems to agree, “Scaling LLMs as they keep scaling will start to lose their efficaciousness.” Eisman is not pulling out of AI stocks. He still owns key players such as Nvidia, Microsoft and Meta. “I own a lot of these stocks too. I own them all,” he told on CNBC. Eisman’s former Big Short counterpart, Michael Burry, is taking the worry even further. For months, Burry has argued that AI demand is overstated and recently shorted major AI names including Nvidia and Palantir. Both men agree on one thing: nobody yet knows whether the enormous AI spending wave will generate enough returns to justify its size. Eisman put out the entire debate into a single line, “The big question on AI spending is, what kind of returns and cost savings the massive investments will bring.” As of 15th December 2025, 10:20PM, Alphabet’s Class C shares (GOOG) are trading at $307.52, which is about $3 lower than the previous close. Class A shares (GOOGL) are at $306.62, also down by around $2.67. Meta stock opened at $645.43 and traded between $638.70 and $650.96 during the day. About 5.47 million shares were traded. Over the past year, the stock has gone as high as $796.25 and as low as $479.80.