Key Takeaways
- • Citadel founder Ken Griffin stated that investors have "almost certainly" inflated a bubble in the artificial intelligence market, according to MarketWatch, speaking at a Citadel conference in New York.
- • Griffin warned that an oversupply of GPUs could drive down computing power prices and challenge the economic model of companies investing in AI infrastructure, drawing an analogy to airlines cutting fares on unfilled seats.
- • Griffin cited the early-2000s telecom boom, when excessive fiber-optic construction cut communication costs, as a precedent for the risk.
- • Griffin does not view a bubble as the end of AI, noting that firms surviving such downturns often become market leaders for decades, as Amazon did after the internet boom.
- • Microsoft, Meta, Oracle, Amazon, and Alphabet collectively carried about $1.09 trillion in future lease obligations, mainly for AI data centers, as of August, according to Reuters.
Summary
Citadel founder Ken Griffin has said that investors have "almost certainly" inflated a bubble in artificial intelligence, according to MarketWatch. Speaking at a Citadel conference in New York, the billionaire warned that an oversupply of graphics processing units (GPUs) could push down the price of computing power and undermine the economics of companies investing in AI infrastructure.
To illustrate the risk, Griffin compared the market to the airline industry: when supply exceeds demand, airlines must lower fares on unfilled seats. Similarly, he argued, if computing capacity outstrips market needs, GPU rental prices could fall sharply. He pointed to the telecommunications boom of the early 2000s, when excessive fiber-optic network construction led to a steep drop in communication costs.
Griffin does not, however, see a bubble as the end of the AI industry. He noted that companies that survive such downturns often become dominant market leaders for decades, pointing to the emergence of tech giants such as Amazon after the internet boom.
Griffin is not alone in his concerns. In August, Bridgewater Associates founder Ray Dalio compared the current AI enthusiasm to the bubbles of 1929 and 2000, citing high company valuations, speculative investment, and the use of leverage. The Bank for International Settlements has also warned of the risks of excessive AI infrastructure investment, noting that competition for leadership may push companies to spend beyond what potential returns justify. The BIS flagged additional risks including debt financing and circular relationships in which companies invest in AI developers and then sell them equipment and computing power.
Balance-sheet exposure is significant. As of August, Microsoft, Meta, Oracle, Amazon, and Alphabet collectively held about $1.09 trillion in future lease obligations, primarily for AI data centers, according to Reuters.
To illustrate the risk, Griffin compared the market to the airline industry: when supply exceeds demand, airlines must lower fares on unfilled seats. Similarly, he argued, if computing capacity outstrips market needs, GPU rental prices could fall sharply. He pointed to the telecommunications boom of the early 2000s, when excessive fiber-optic network construction led to a steep drop in communication costs.
Griffin does not, however, see a bubble as the end of the AI industry. He noted that companies that survive such downturns often become dominant market leaders for decades, pointing to the emergence of tech giants such as Amazon after the internet boom.
Griffin is not alone in his concerns. In August, Bridgewater Associates founder Ray Dalio compared the current AI enthusiasm to the bubbles of 1929 and 2000, citing high company valuations, speculative investment, and the use of leverage. The Bank for International Settlements has also warned of the risks of excessive AI infrastructure investment, noting that competition for leadership may push companies to spend beyond what potential returns justify. The BIS flagged additional risks including debt financing and circular relationships in which companies invest in AI developers and then sell them equipment and computing power.
Balance-sheet exposure is significant. As of August, Microsoft, Meta, Oracle, Amazon, and Alphabet collectively held about $1.09 trillion in future lease obligations, primarily for AI data centers, according to Reuters.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.