The Big Short's Steve Eisman Warns Of AI's 2 Big Risks In 2026: Power Constraints, Diminishing Returns On Larger LLMs

Investor Steve Eisman, famed for his role in predicting the 2008 financial crisis, which was portrayed in the 2015 movie “The Big Short,” has a bold prediction on the AI boom as we head into the new year.

AI’s Two Big Risks In 2026

On the Real Eisman Playbook podcast on Saturday, Eisman stated that two major risks could derail the AI momentum in 2026: growing power shortages and diminishing returns from scaling large language models (LLMs).

“The current binding constraint on AI growth is power,” he said. “You can build an AI data center, but if you can't plug it in, what good is it?” while citing two recent examples of data centers in Santa Clara, California, which have no electricity, amid massive shortfalls.

See Also: AI Adoption Isn’t Driven By Executives, But By Workers: What 100K Users Reveal

While U.S. power demand is now growing at roughly 3% annually, he warned that new energy infrastructure takes time to build. “From drawing up the plans to completion, it's at least a two to three-year process,” he said.

More concerning, Eisman argued, is an intellectual risk tied to the industry's reliance on ever-larger language models, which he said could be “a dead end.”

Eisman highlighted OpenAI’s release of ChatGPT 5.0, which critics had pointed out was no better than its earlier version, ChatGPT 4.0. “Since then, there has been a growing but still minority opinion that believes that continuing to scale LLMs is becoming a dead end.”

“If this opinion gains further traction,” Eisman warned, “the need for more chips will diminish and tech hardware growth will slow across the board and the virtuous cycle will reverse.”

According to Eisman, “the potential realization that the ever larger LLMs won't get us to the finish line,” which is AGI, or artificial general intelligence, “is the fundamental risk to the entire AI story.”

Hype Has Peaked

In JPMorgan Chase & Co.’s (NYSE:JPM) 2026 outlook, a survey of CEOs shows that the market is now recalibrating expectations when it comes to AI, with executives now pressing for clear returns on investment, as opposed to blind experimentation.

However, leading tech analyst Gene Munster of Deepwater Asset Management has said that the AI trade is “still early,” with 2026 being year three of a “5-year bull market.”

Driven by the AI-trade, Munster expects the Nasdaq to end 2026 up by 10% or more, while adding that the “current bubble talk is actually constructive because it keeps expectations in check and lowers the risk of an actual bubble forming.”

Read More:

• Google DeepMind CEO Warns AI Startup Bubble Is Forming As Early-Stage Firms Raise Tens Of Billions Before Launch

Photo courtesy: Shutterstock