Ken Griffin Sees AI Bubble; GPUs Compared to Airline Seats

Ken Griffin Says AI Is “Almost Certainly” in a Bubble—and Compares GPUs to Airline Seats

Ken Griffin, founder and CEO of Citadel, said investors are “almost certainly” in an artificial intelligence bubble, warning that the economics of the AI chip market could set the stage for sharp price compression if supply outruns demand.

Speaking on Oct. 20 at Citadel’s Future of Global Markets Conference in New York, Griffin argued that the key risk isn’t simply enthusiasm around AI, but the possibility that the market for graphics processing units (GPUs)—the chips widely used to train and run AI models—could start behaving like the airline industry.

Griffin compared GPUs to airline seats because both have a very low marginal cost to serve one more customer. If a plane has empty seats, even a deeply discounted ticket still brings in revenue. In Griffin’s view, if data centers and cloud providers end up with more GPU capacity than the market needs, pricing could “go into freefall,” undermining the assumptions behind current investment and spending plans.

His comments come as some Wall Street observers debate whether the AI cycle is starting to cool. BTIG, for example, has warned of a prolonged semiconductor slump that could stretch into 2027, after the Philadelphia Semiconductor Index (SOX) posted its largest weekly drop since April 2025.

Griffin also questioned how long today’s high GPU economics can last. He noted that chips that are five or six years old still earn what he described as extraordinary rates, but said he is unsure that will remain true over the next five or six years.

Beyond hardware, Griffin said corporate buyers are still working out where premium AI models are worth the cost and where cheaper open-source alternatives can do the job—dynamics that could constrain spending at the same time AI labs are investing heavily in infrastructure.

Even with the bubble warning, Griffin did not frame the situation as a simple call to abandon the sector. He said transformative economic shifts are often intertwined with bubbles, adding that the companies that survive typically become market leaders for decades.

Griffin’s own view of AI has shifted over time. He has described moving from dismissive earlier in 2026 to “unsettled” by the technology’s capabilities by May, citing agentic AI tools that cut some Citadel research processes from six to eight weeks down to just a few hours.

Not everyone accepts the dot-com-era comparison. Yardeni Research President Ed Yardeni has argued the late-1990s bubble analogy no longer fits and has forecast a higher S&P 500 by mid-2027.

In the market, the immediate reaction appeared muted. Nvidia shares finished the session down 0.29% at $216.47 before rising 2.12% in after-hours trading to $221.06. The Invesco QQQ Trust slipped 0.10% to $717.63.

• What happened: Ken Griffin said investors are “almost certainly” in an AI bubble at a Citadel conference in New York.
• Why it matters: He warned that a GPU supply glut could cause prices to collapse, challenging the financial assumptions behind AI infrastructure buildouts.
• Broader context: Even if bubbles form, Griffin said the survivors often become long-term leaders—while debate continues over whether current AI conditions resemble past tech booms.