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

Ken Griffin, the founder and chief executive officer of Wall Street trading and hedge fund giant Citadel, believes that the current fervor surrounding artificial intelligence is "almost certainly" a market bubble. However, despite the towering valuations and intense capital expenditures defining the tech sector, the billionaire financier is not advising investors to bail on their AI holdings.

Speaking at Citadel’s Future of Global Markets Conference in New York, Griffin laid out a nuanced perspective on the historic technology boom that has dominated global equities over the past several years. While asset prices have undeniably outpaced the immediate fundamental earnings of the underlying businesses—the textbook definition of a speculative bubble—Griffin emphasized that historical precedent suggests such phases are a necessary incubator for world-changing innovation.

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

This dynamic carries substantial implications that stretch far beyond traditional technology-focused investors. Because the massive infrastructure buildout surrounding artificial intelligence has served as the primary pillar supporting broader equity markets over the last year, a vast number of mainstream index funds, institutional portfolios, and everyday retirement accounts now hold a meaningful slice of the AI trade. A market bubble typically materializes when asset prices climb far faster than the operational realities of the companies behind them can justify, creating a tense waiting period until fundamental reality eventually catches up.

Yet, from a historical perspective, Griffin argues that the emergence of a bubble is hardly catastrophic. Drawing comparisons to the transformative internet boom of the late 1990s—noting with hindsight that he wishes he had held onto his Amazon stock—he pointed out that nearly every radical structural transformation of the global economy has been intricately intertwined with a speculative phase. The corporate entities that successfully navigate the turbulence and emerge on the other side tend to dominate their respective markets for decades, even if short-term valuations experience severe corrections.

Griffin’s own firm serves as a prime case study of the profound operational shifts driven by the technology. He revealed that Citadel has successfully deployed advanced AI agents—sophisticated software systems capable of executing complex, multi-task workflows autonomously—to streamline internal operations, dramatically cutting down the time required for high-level proprietary research from weeks to mere hours.

At the same time, Griffin offered a sharp structural critique of the economics powering the semiconductor industry. He compared graphics processing units (GPUs), the specialized hardware essential for training and running massive AI models, to airline seats. Both commodities share a very low marginal cost to run once the initial capital expenditure is spent, meaning the incremental cost of serving one additional customer is minuscule. Consequently, when production capacity ultimately outstrips real-world demand, prices face the risk of a sharp collapse.

To illustrate this vulnerability, Griffin pointed back to the telecommunications boom around the year 2000, a period when aggressive capital spending led companies to lay far more fiber-optic cable than immediate market demand warranted, ultimately causing communication prices to plummet. He also expressed skepticism that older-generation chips—hardware sitting five or six years past its prime that currently commands extraordinary rental rates—will manage to sustain such pricing power for another half-decade.

Corporate buyers are already growing increasingly discerning, Griffin observed. Businesses are actively evaluating whether complex tasks truly justify the premium subscription models offered by leading labs like OpenAI and Anthropic, or if they can adequately be handled by cheaper open-source alternatives that organizations can download and run on their own local infrastructure. Corporate America’s ability to extract tangible financial value from artificial intelligence remains uneven, with executives increasingly reviewing steep software and cloud bills and asking fundamental questions about where the promised productivity gains are materializing.

Market forecasts underscore the staggering scale of the financial commitments currently flowing into the sector. Gartner has projected that worldwide spending on artificial intelligence will reach nearly $1.5 trillion, encompassing everything from advanced semiconductors and enterprise servers to AI-enabled mobile devices. Meanwhile, Nvidia CEO Jensen Huang noted at Davos that the broader industry is currently only "a few hundred billion dollars" into what he described as the largest infrastructure buildout in human history, with trillions more in capital investment still required. If end-users begin to balk at premium pricing, the hardware manufacturers and cloud computing giants relying on that continuous capital expenditure cycle would feel the shockwaves first.

The high-stakes nature of this market dynamic has already triggered dramatic fallout in the hedge fund sector. On July 30, Situational Awareness—a specialized fund founded by former OpenAI researcher Leopold Aschenbrenner—offloaded the vast majority of its public stock portfolio to Citadel at roughly a 10% discount. The transaction followed a severe drawdown that reportedly saw the fund’s portfolio value plunge by about 67% over the course of July.

Situational Awareness had aggressively utilized leverage, borrowing as much as four times its underlying capital to fuel its purchases of high-flying AI stocks. As asset prices declined, institutional lenders repeatedly issued margin calls, demanding additional collateral on a daily basis. Forced selling—a scenario where a fund is compelled to liquidate assets due to creditor demands rather than strategic choice—tends to create a self-reinforcing downward spiral, as each wave of liquidations depresses market prices further and triggers subsequent rounds of margin calls.

Citadel moved decisively to absorb that risk, and by August, Griffin confirmed in a client letter that the firm had successfully shed more than 80% of the aggregate risk associated with the acquired portfolio. In effect, the prominent Wall Street executive warning of an AI bubble was also positioned to acquire discounted assets during a moment of sector-wide distress.

Broader macroeconomic pressures continue to loom over the financial markets as well. Earlier in the year, Griffin issued warnings regarding potential geopolitical risks, noting that a prolonged closure of the Strait of Hormuz—a vital global oil shipping lane—could drive energy prices significantly higher and push the global economy into a recession.

Meanwhile, Wall Street analysts are beginning to factor a cooling period into their forward projections. According to research published by Goldman Sachs, capital expenditure growth among the largest AI infrastructure builders is widely expected to decelerate markedly, slowing from an estimated 75% year-over-year expansion in the third quarter of 2025 down to roughly 25% by the end of 2026. As the market transitions from the initial phase of unbridled infrastructure accumulation to a more rigorous phase of financial accountability, the durability of the AI boom will increasingly depend on whether corporate deployment can match the monumental sums invested.