Ken Griffin Warns AI May Be in a Bubble

Highlights

Citadel founder Ken Griffin said on Oct. 20 that investors are “almost certainly” in an AI bubble, while arguing that companies emerging from such periods can become market leaders for decades. He compared GPUs with airline seats: both can be inexpensive to supply at the margin, making prices vulnerable if capacity exceeds demand. Griffin also said corporate customers are scrutinizing AI bills and weighing premium OpenAI and Anthropic models against cheaper open-source alternatives. His view combines confidence in AI’s long-term impact with caution about valuations, chip supply, and the returns companies can justify.

Sentiment Analysis

• The article’s sentiment is mixed, with a cautious emphasis. Griffin’s description of AI investment as “almost certainly” a bubble conveys concern about elevated expectations and the possibility that prices could fall if supply exceeds demand.
• The account is not wholly bearish: Griffin says companies that survive bubbles may lead markets for decades and describes AI agents as having sharply shortened some research tasks at Citadel. These points suggest meaningful long-term economic potential.
• Discussion of buyers comparing premium models with open-source alternatives, and of companies asking where productivity gains are, adds uncertainty about whether current spending can be sustained. The account of a leveraged fund facing lender demands illustrates how falling prices can intensify selling.
• Overall, the article presents a neutral-to-cautious assessment rather than a prediction of an imminent collapse. It distinguishes the technology’s possible lasting value from the financial risks of overbuilding and excessive expectations.

Article Text

Ken Griffin, founder of Citadel, said on Oct. 20 that investors are “almost certainly” in an artificial-intelligence bubble. Speaking at Citadel’s Future of Global Markets Conference in New York, he nevertheless stopped short of urging investors to sell their AI holdings or predicting that the market would collapse. Griffin argued that bubbles can leave behind technologies and businesses that transform the economy, and that the companies that endure may become market leaders for decades.

The distinction between a technology’s lasting importance and the price paid for its shares is central to Griffin’s comments. A bubble occurs when asset prices rise faster than the businesses’ underlying prospects can support. Because AI infrastructure investment has helped underpin equity markets over the past year, the stakes extend beyond technology specialists: index funds and retirement accounts may also have exposure. Griffin recalled the internet boom and said he wished he had retained his Amazon shares, using that experience to illustrate how a bubble can cause losses while still producing companies with lasting influence.

Griffin also described practical uses of AI within Citadel. AI agents—software capable of carrying out multistep tasks independently—have reduced some high-level research work at the firm from weeks to a couple of hours, he said. The example offers a specific account of productivity gains, although it does not establish how broadly similar results can be achieved across businesses. Griffin said corporate America has been uneven in extracting value from AI, with companies reviewing bills from AI labs and asking where the productivity benefits are.

One potential source of financial risk is the economics of computing hardware. Griffin compared graphics processing units, or GPUs, used to train and run AI models with airline seats. In both cases, the marginal cost of serving one additional customer can be low. If companies build more capacity than customers need, providers may compete by cutting prices. Griffin pointed to the telecom buildout around 2000, when excess fiber-optic cable capacity was followed by falling communication prices.

He questioned whether chips that are five or six years old will continue earning what he called extraordinary rental rates for another five or six years. The concern is relevant to businesses and investors counting on continued demand for computing capacity. At the same time, corporate buyers are sorting tasks by what they require: some may justify premium models from OpenAI and Anthropic, while others may be handled by cheaper open-source models that customers can download and run on their own computers. A shift toward lower-cost alternatives could put pressure on the revenues expected by AI labs and infrastructure suppliers.

The scale of planned investment is considerable. Gartner forecasts worldwide AI spending at nearly $1.5 trillion for 2025, including chips, servers, and AI-enabled phones. Nvidia CEO Jensen Huang said at Davos in January that the industry was only “a few hundred billion dollars” into what he described as the largest infrastructure buildout in human history, with trillions still to build. These estimates reflect expectations for continued growth, but the article notes that customers’ willingness to pay premium prices will affect companies relying on the spending.

The article also recounts a case of financial strain in AI-related investments. 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. As prices declined, lenders requested additional collateral each day. When investors must sell to meet lenders’ demands, the resulting forced selling can push prices lower and prompt further demands.

In an August letter to clients, Griffin wrote that Citadel had since shed more than 80% of the risk in that portfolio. The detail complicates a simple interpretation of his bubble warning: Citadel also acquired the discounted stocks, then reduced its exposure. Separately, Griffin had warned earlier in the year that a Strait of Hormuz closure lasting another six to 12 months would materially raise energy prices and push the world into recession. Goldman Sachs Research noted in December that Wall Street analysts expected capital-spending growth among the largest AI infrastructure builders to slow from 75% in the third quarter of 2025 to 25% by the end of 2026. Together, these points frame Griffin’s position as neither an outright rejection of AI nor an assurance that current investment levels can continue. The central uncertainty is whether lasting AI productivity will support the enormous spending and prices associated with the current buildout.

Key Insights Table