Michael Burry, the celebrated investor who foresaw the housing market collapse ahead of 2008, has shifted his analytical gaze towards a current market favorite: artificial intelligence. Although Burry has deregistered his hedge-fund firm, Scion Asset Management, from routine regulatory disclosures, he continues his investment activities, identifying potential mispricings in the market.
At the heart of Burry's focus is Phil Clifton, Scion's former associate portfolio manager, whose analyses lend a critical eye to the AI industry's rapid infrastructure expansion. Clifton contends that despite the surge in generative AI adoption, the underlying economics do not justify the extensive infrastructure investments the industry is witnessing.
In an October farewell letter to Scion investors, Burry lauded Clifton as the "most prodigious thinker" he has ever known. CNBC obtained several of Clifton's research documents, penned before he ventured out to establish his own firm, Pomerium Capital. These documents form the foundation of Scion's skeptical stance on the AI market.
Clifton argues that the financial community is overestimating the economic impact of AI technology, writing, "Just because a technology is good for society or revolutionizes the world doesnât mean that itâs a good business proposition."
Low Margins
While AI appears to be widely adoptedâmore than 60% of U.S. adults reportedly interact with AI at least several times a week, per Pew Research CenterâClifton points out that its economic scale on the demand side is "surprisingly small." OpenAI, a leading entity in the AI space, is poised to achieve over $20 billion in annualized revenue this year. However, this figure pales compared to the colossal scale of AI infrastructure investments. Hyperscalers have quadrupled their capital expenditures recently, nearly reaching $400 billion annually, and expectations predict a $3 trillion spend over the next five years, according to the Man Group.
"We assume other generative AI services in aggregate are insufficient to justify the sums being spent on infrastructure," Clifton observed.
Historyâs Warnings
Burry draws parallels to the early-2000s telecom surge, where investment in fiber-optic networks vastly outstripped actual use, leading to an overcapacity dilemma in the U.S. The resultant market suffered a drastic 70% fall in wholesale telecom pricing over a single year, as Scion noted.
Clifton theorizes that cloud giants are now in a similar race, heavily investing in AI infrastructure with the hope that demand will eventually align. Should widespread AI adoption take longer than anticipated, the financial feasibility of these expansive data center agreements could be compromised.
There are already signs of hesitancy among major tech companies. Microsoft has canceled data center projects in the U.S. and Europe that were supposed to consume 2 gigawatts of electricity, citing an excess of supply. Meanwhile, Alibaba's chairman has sounded alarm bells over a burgeoning bubble in AI infrastructure.
The Nvidia Exposure
Nvidia has been a major beneficiary of AI-related expenditures, with an impressive surge in stock value fueled by unprecedented GPU orders from cloud providers. However, Scion questions whether these customers will eventually reap economic returns from their investments.