Ken Griffin Says AI Is ‘Almost Certainly’ in a Bubble—And Compares GPUs to Airline Seats

Ken Griffin, founder of the Wall Street giant Citadel, said Tuesday that investors are “almost certainly” in an AI bubble—though he isn’t telling them to dump their AI stocks.

“What’s left in the wake of these bubbles, if we want to use that word, is often game changing for the world. And the companies that are on the other side of that bubble are the market leaders for decades to come,” Griffin said at Citadel’s Future of Global Markets Conference in New York on Oct. 20

This reaches beyond tech investors, since the AI buildout has underpinned equity markets over the past year, so many index funds and retirement accounts hold a slice of it. A bubble is what happens when prices climb far faster than the businesses behind them can justify, until reality catches up. Citadel is a hedge fund, a private firm that invests money for wealthy clients and institutions.

But the fact that it seems like a bubble isn’t that big a deal if you look at the big picture, according to Griffin. “Any radical transformation of our economy is generally intertwined with a bubble,” he said, recalling the internet boom and adding that he wishes he had kept his Amazon stock.

The companies that emerge on the other side of a bubble tend to lead markets for decades, he argued, though he stopped short of predicting a collapse. His own firm is a case study: Griffin said AI agents, software that carries out multistep tasks on its own, cut some high-level Citadel research from weeks to a couple of hours.

Griffin compared GPUs, the graphics chips that many companies use to train and run AI models, to airline seats. Both have a very low marginal cost to run, meaning the extra expense of serving one more customer is tiny, so prices can collapse once supply outstrips demand.

He pointed to the telecom boom around 2000, when companies laid more fiber-optic cable than demand required and communication prices fell. He also doubted that chips five or six years old, which still earn rental rates he called extraordinary, will keep doing so for another five or six years.

Buyers are getting pickier, Griffin said, sorting which jobs justify premium models from OpenAI and Anthropic and which can run on cheaper open-source alternatives, meaning models anyone can download and run on their own computers. Corporate America has been uneven at extracting value from AI, he added, with companies studying their bills from the labs and asking where the productivity is.

Gartner forecasts worldwide AI spending at nearly $1.5 trillion for 2025, counting everything from chips and servers to AI-enabled phones. Nvidia CEO Jensen Huang said at Davos in January that the industry is only “a few hundred billion dollars” into what he called the largest infrastructure buildout in human history, with trillions still to build. If customers balk at premium prices, the chip and cloud companies counting on that spending would feel it first.

On July 30, Situational Awareness, a hedge fund founded by former OpenAI researcher Leopold Aschenbrenner, sold most of its public stock portfolio to Citadel at roughly a 10% discount after its value reportedly fell about 67% in July.

The fund had borrowed as much as four times its capital to buy AI stocks, a tactic called leverage, and its lenders asked for more collateral every day as prices fell. Selling because lenders demand it rather than by choice is known as forced selling, and it feeds itself because each sale pushes prices down and triggers more demands.

Citadel has since shed more than 80% of the risk in that portfolio, Griffin wrote in an August letter to clients, which means the man calling the bubble also bought the discounted stocks.

Griffin warned earlier this year that if the Strait of Hormuz, a key oil shipping lane, stayed closed for another six to 12 months, energy prices would rise materially and push the world into recession.

Wall Street analysts expect growth in capital spending, the money poured into chips and data centers, by the largest AI infrastructure builders to slow from 75% in the third quarter of 2025 to 25% by the end of 2026, Goldman Sachs Research noted in December.