Citadel's Ken Griffin Warns AI Chip Glut Could Match Airline Seat Pricing Plunge

Citadel CEO Ken Griffin said an oversupply of AI chips could send prices into freefall, drawing a parallel between graphics processing units (GPUs) and airline seats that still earn money when sold cheaply.

Griffin said AI almost certainly contains bubbles, though he stopped short of predicting a collapse. Businesses that make it through to the other side of a bubble are the ones that lead markets for decades, he added.

Does Ken Griffin's Airline Comparison Work for AI Chips?

Griffin made the comments during a conversation with Bloomberg's Dani Burger at Citadel's annual Future of Global Markets Conference in New York.

“GPUs like airlines have a very low marginal cost to run.”

Ken Griffin made those comments to Bloomberg.

An unfilled seat on an airplane can be sold for as little as $20 and still be profitable, he said, so fares collapse once capacity outruns demand.

Griffin pointed to the telecom build-out around 2000, when more fiber was laid than demand justified and communication prices dropped.

If the supply of GPUs runs ahead of demand, pricing could go into freefall, he said, overturning the investment assumptions behind today's AI buildout.

What Is Behind Wall Street's Doubts About AI Hardware Returns?

Griffin framed the recent AI-related selloff as investors doing their homework, adding that hedge fund deleveraging should be the 10th explanation anyone reaches for.

BTIG, a Wall Street firm, now expects an extended chip slump that could run into 2027, after the Philadelphia Semiconductor Index (SOX) lost 11.4% in a week, its biggest weekly fall since April 2025.

Griffin also raised questions about how long a GPU remains useful. Chips that are five or six years old still command rates he described as extraordinary, but he doubted that would hold for another five or six years.

Corporate buyers are also deciding which tasks deserve the premium models offered by OpenAI and Anthropic and which can be served by cheaper open-source alternatives, Griffin said, and he agreed that this puts pressure on the labs' spending.

The dot-com comparison is not one every strategist embraces, though. Ed Yardeni, president of Yardeni Research, predicts the S&P 500 at 8,400 by mid-2027, arguing that the 1999 and 2000 bubble analogy no longer applies.

The deciding factor is whether revenue from AI can catch up with the capacity under construction, since that will determine how much of the spending earns a return.