Ken Griffin Says AI Is Almost Certainly in a Bubble, as Citadel Buys Assets From a Shrunken AI Fund
Citadel founder Ken Griffin said investors are "almost certainly" in the midst of an AI bubble, according to a Seeking Alpha report published October 21, 2026. The same report says he also described companies as uncertain about the returns on their AI investment.
Griffin has used bubble language about AI before. In a September 25, 2025 CNBC interview with Sara Eisen, he said there were "obviously echoes of the dotcom bubble" in the current AI moment. His July 2026 remarks went further. A YouTube video carrying the title of a Goldman Sachs Exchanges interview has him saying the AI bubble is "much worse" than the dot-com bubble, and that agentic AI, software that carries out multistep tasks on its own, is expected to grow. Goldman Sachs published that interview on July 9, 2026. In it, Griffin discussed the rising cost of AI compute, which is the processing power needed to train and run AI models, and geopolitical vulnerabilities, including US data centers and US-China tensions.
Market coverage has picked up the warning. A MarketWatch quote page for Super Micro Computer, dated October 20, 2026, carried the headline "Ken Griffin Warns of AI Bubble Risks" next to "AI Stocks Jump On SpaceX, Supermicro."
In my view, that side-by-side placement is the most telling detail. A sharp warning from a prominent allocator and a rally in the names most tied to AI hardware landed on the same day, which suggests the market is treating the warning as one more input rather than a reason to stop buying.
Griffin's caution has not led him to step back from the sector. A MarketWatch Key Words item reports that he warned investors not to chase stocks higher while also explaining why he is investing in AI anyway.
The broader context here is that both positions can be held at once. A bubble call is a claim about price relative to cash flows that have yet to arrive, and it says little about timing. A manager can believe valuations are stretched and still hold exposure.
CNBC reported on July 31, 2026 that Leopold Aschenbrenner's AI hedge fund, Situational Awareness, had fallen from about $45 billion to around $10 billion in assets on falling prices. The same day, CNBC reported that Citadel had reached a deal to buy the publicly traded assets of the collapsed fund.
In my view, the fund is the more concrete data point than the quote. A fall of that scale, driven by falling prices, is first a markdown problem, meaning the fund's holdings are worth less on paper, and only then a question about investors pulling money out. The firm whose founder is now warning about AI valuations is also the buyer of assets from a fund whose AI-focused book lost roughly three-quarters of its size in the reported period. For anyone holding AI-exposed positions, directly or through funds, the relevant question is how much of their return depends on the same expectations Griffin is calling into question. That is a question for each investor to work through with their own adviser.
The bubble framing also deserves some skepticism. "Bubble" is a judgment that prices have run ahead of what the underlying businesses can earn, and that judgment stays contestable until the earnings either arrive or do not. Griffin's statements, as reported across 2025 and 2026, have moved from "echoes" of the dot-com era to "much worse," which signals rising concern but supplies no timeline. A forecast without a date cannot be wrong on a schedule, and it cannot be checked in the short run.
A prominent hedge fund founder has said, in reported remarks, that investors are almost certainly in an AI bubble and that companies are unsure what their AI spending will return. A large AI-focused fund has shrunk sharply on falling prices, and Citadel has agreed to buy its public assets. Everything beyond that is interpretation, and readers should weigh it as such.
