Ken Griffin Says AI Is ‘Almost Certainly’ in a Bubble, But Advises Investors to Hold On

Ken Griffin, the billionaire founder and CEO of Wall Street giant Citadel, believes that investors are "almost certainly" operating inside an artificial intelligence bubble—yet he is firmly advising them against dumping their tech holdings.

Speaking at Citadel’s Future of Global Markets Conference in New York on October 20, Griffin offered a pragmatic and historical perspective on the current market mania surrounding artificial intelligence. While recognizing the classic symptoms of an overextended market, he emphasized that speculative bubbles have historically accompanied the most profound economic shifts in modern history.

"What’s left in the wake of these bubbles, if we want to use that word, is often game changing for the world," Griffin told the audience. "And the companies that are on the other side of that bubble are the market leaders for decades to come."

The ripple effects of this debate reach far beyond dedicated tech investors. Because the massive infrastructure buildout surrounding artificial intelligence has heavily underpinned broader equity markets over the past year, countless index funds, mutual accounts, and retirement portfolios hold a direct slice of the sector. A market bubble inherently materializes when asset prices climb far faster than the underlying businesses can fundamentally justify, creating a tense environment where market valuations race ahead until reality eventually catches up. Citadel itself operates as a premier hedge fund, a private firm managing capital for wealthy individuals and massive institutional clients.

Despite the obvious parallels to past market manias, Griffin argues that the presence of a bubble should not trigger panic among long-term investors. Recalling the explosive internet boom of the late 1990s, he candidly admitted that looking back, he wishes he had held onto his early Amazon stock rather than retreating at the first sign of irrational exuberance. Any radical transformation of the global economy, he noted, is generally intertwined with a speculative bubble.

The companies that successfully navigate the turbulence and emerge on the other side tend to dominate global markets for decades, he argued. Even so, the Citadel founder stopped short of predicting an imminent and catastrophic market collapse.

In fact, Citadel’s own internal operations serve as a living case study of the immense, transformative power driving the very technology he labels a bubble. Griffin revealed that advanced AI agents—sophisticated software capable of carrying out complex, multistep research tasks completely on their own—have dramatically streamlined workflows within his own firm, cutting down high-level proprietary research projects that once took weeks into a matter of a couple of hours.

Yet, enthusiasm for the technology must be weighed against economic realities, particularly regarding hardware. Griffin drew a direct economic parallel between Graphics Processing Units (GPUs)—the specialized semiconductor chips that countless companies rely on to train and run massive AI models—and airline seats. Both commodities share a distinct economic profile defined by a very low marginal cost to run. Because the incremental expense of serving one additional customer or running one more workload is microscopic, prices can plummet precipitously the moment that supply outstrips consumer demand.

To illustrate this vulnerability, he pointed back to the telecommunications boom around the year 2000. During that era, aggressive telecom companies laid vastly more fiber-optic cable than actual market demand required, leading to a massive oversupply and a subsequent collapse in communication prices. He also expressed skepticism that older generations of AI chips, which currently still command rental rates he described as extraordinary, will manage to maintain those premium pricing levels for another five or six years.

Corporate buyers are already growing considerably more discerning, according to Griffin. Businesses are actively sorting out which specialized tasks genuinely justify paying premium subscription models to labs like OpenAI and Anthropic, and which workloads can easily and cost-effectively run on cheaper open-source alternatives—powerful models that anyone can download and run independently on their own local hardware. Furthermore, corporate America’s track record in successfully extracting measurable value from artificial intelligence has been notably uneven. Many companies are currently reviewing steep billing statements from AI labs and asking fundamental questions about where the promised productivity gains are actually materializing.

Independent industry projections highlight the sheer scale of the financial commitments at stake. Market research firm Gartner has forecasted that worldwide artificial intelligence spending will approach a staggering $1.5 trillion for 2025 alone, encompassing everything from advanced semiconductors and dedicated servers to AI-enabled consumer devices. Meanwhile, Nvidia CEO Jensen Huang noted during remarks at Davos in January that the technology sector is still only "a few hundred billion dollars" into what he characterized as the largest infrastructure buildout in human history, with trillions of dollars in additional investments required ahead. Should corporate customers eventually balk at premium hardware and software prices, the chip manufacturers and cloud infrastructure providers relying heavily on that spending would feel the shockwaves first.

The high-stakes nature of this market dynamic was vividly illustrated this past summer. On July 30, Situational Awareness—a specialized hedge fund founded by former OpenAI researcher Leopold Aschenbrenner—was forced to sell the vast majority of its public stock portfolio directly to Citadel at roughly a 10% discount. The distress sale occurred after the fund’s portfolio value reportedly plunged by about 67% over the course of July.

Situational Awareness had aggressively utilized financial leverage, borrowing as much as four times its total capital base to load up on high-flying AI stocks. As market prices retreated, its institutional lenders repeatedly issued margin calls, demanding additional collateral on a daily basis. Selling assets out of sheer contractual necessity rather than strategic choice is known as forced selling—a self-reinforcing downward spiral where each wave of liquidations pushes market prices lower, triggering fresh collateral demands and compounding the distress.

In an ironic twist, Citadel stepped in to absorb the risk. Griffin wrote in an August letter to clients that his firm had successfully shed more than 80% of the aggregate risk associated with that acquired portfolio, demonstrating that even while publicly warning about the existence of an AI bubble, the firm was actively capitalizing on discounted market dislocations.

Griffin’s macroeconomic outlook extends well beyond technology sectors. Earlier this year, he warned that if the Strait of Hormuz—a critically important global oil shipping lane—remained closed for an extended period of six to 12 months, energy prices would rise materially and threaten to push the global economy into a recession.

As the market continues to absorb these competing pressures, Wall Street financial analysts anticipate a cooling off in capital expenditures. According to research published in December by Goldman Sachs Research, capital spending—the massive sums of money poured directly into high-end chips and sprawling data centers by the largest AI infrastructure builders—is expected to slow down noticeably, shifting from a blistering 75% growth rate in the third quarter of 2025 down to a more moderate 25% by the end of 2026.