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

Ken Griffin, the founder of Citadel, one of Wall Street's major financial firms, expressed his views on the current state of artificial intelligence (AI) investments during a conference in New York on October 20. He stated that investors are "almost certainly" experiencing a bubble in the AI sector, but he is not advising them to sell their AI-related stocks. Griffin emphasized that significant transformations in the economy often coincide with such bubbles, recalling the internet boom and expressing regret for not holding onto his Amazon shares.

During the Future of Global Markets Conference, Griffin noted that the recent surge in AI investment has influenced equity markets broadly, affecting many index funds and retirement accounts. He explained that a bubble occurs when asset prices rise much quicker than the underlying companies can support, ultimately leading to a correction when reality sets in. Citadel, as a hedge fund, manages investments for affluent clients and institutions, making the implications of these market trends far-reaching.

Despite the signs of a bubble, Griffin maintained that this should not be a cause for alarm when viewed from a broader perspective. He pointed out that the companies that successfully navigate the aftermath of a bubble often become market leaders for extended periods. While he refrained from making predictions about a potential market crash, he shared insights from Citadel's own experiences, highlighting how AI agents have significantly reduced the time required for complex research tasks from weeks to just a few hours.

Griffin made an analogy between graphics processing units (GPUs) — essential for training and operating AI models — and airline seats. He explained that both have low marginal costs, meaning the additional expense of accommodating one more user is minimal. This dynamic can lead to price drops when supply exceeds demand. He referenced the telecom boom of the early 2000s, where excessive investment in fiber-optic cable resulted in a significant decline in communication prices.

He also expressed skepticism about the longevity of high rental rates for older chips, which he described as "extraordinary" but unlikely to sustain for another five or six years. According to Griffin, buyers are becoming more selective, distinguishing between which tasks necessitate premium models from leading AI firms like OpenAI and Anthropic and which can be handled by more affordable, open-source alternatives that users can operate independently.

Griffin pointed out that corporate America has had mixed success in realizing the value of AI, with many organizations scrutinizing their AI-related expenses while questioning the productivity gains. Looking ahead, Gartner has projected that global spending on AI will reach nearly $1.5 trillion by 2025, encompassing a wide range of technologies from chips and servers to AI-integrated smartphones. Nvidia's CEO, Jensen Huang, mentioned at the Davos conference earlier this year that the industry is only beginning to tap into what he calls the largest infrastructure buildout in human history, with trillions of dollars yet to be invested.

Griffin concluded by noting that if customers hesitate to pay premium prices for AI services, the chip and cloud service providers that are counting on that revenue could be the first to feel the impact. As the landscape of AI investment continues to evolve, the implications for both investors and the broader market remain significant.