Burry's AI 'Big Short' Thesis
Michael Burry, the investor renowned for his prescient call on the 2008 housing market collapse, is now setting his sights on artificial intelligence. He believes the sector, currently a darling of investors, is exhibiting signs of a bubble.
Burry, who recently deregistered Scion Asset Management, continues to actively invest, focusing on perceived mispricings within the market. Key to his AI skepticism is research conducted by Phil Clifton, Scion's former associate portfolio manager. Clifton, who has since launched his own firm, Pomerium Capital, contends that the rapid adoption of generative AI is not supported by sound economics, particularly concerning the massive infrastructure buildout required.
In his farewell letter to Scion investors, Burry lauded Clifton as "the most prodigious thinker" he's ever encountered. Research notes obtained by CNBC outline the core of Scion’s bearish AI thesis, penned before Clifton departed to launch Pomerium Capital.
The Economics of AI Demand
Clifton argues that the investment world is overestimating the economic impact of AI. "Just because a technology is good for society or revolutionizes the world doesn't mean that it's a good business proposition," he wrote.
Despite widespread adoption – Pew Research Center indicates that over 60% of U.S. adults interact with AI multiple times per week – Clifton believes the economic demand remains surprisingly low. He notes that while OpenAI, a market leader, is projected to surpass $2 billion in annualized revenue, this figure is dwarfed by the scale of AI infrastructure investments.
Hyperscalers have reportedly quadrupled their capital expenditure in recent years, approaching $400 billion annually, with projections exceeding $3 trillion over the next five years, according to Man Group. "We assume other generative AI services in aggregate are insufficient to justify the sums being spent on infrastructure," Clifton stated.
Echoes of the Telecom Bubble
Scion draws a parallel between the current AI boom and the telecom bubble of the early 2000s. During that period, substantial investments in fiber-optic networks far exceeded actual usage. Consequently, U.S. capacity utilization plummeted to approximately 5%, and wholesale telecom pricing experienced a sharp decline, falling by roughly 70% in a single year, as noted by Scion.
Clifton suggests that cloud giants are currently engaged in a similar race, expanding AI infrastructure under the assumption that future demand will eventually catch up. However, if widespread AI adoption is slower than anticipated, the economics of these expansive data center deals could become unsustainable.
Signs of Caution Emerge
Some Big Tech companies are already exhibiting signs of hesitation regarding their commitments. Microsoft has reportedly canceled data center projects slated to consume 2 gigawatts of electricity in the U.S. and Europe, citing an oversupply. Alibaba's chairman has also voiced concerns about a potential bubble forming in AI infrastructure.
Nvidia's Role and Depreciation Dilemma
Nvidia has been a primary beneficiary of the AI spending surge, with its stock price soaring alongside unprecedented GPU orders from cloud providers. However, Scion questions whether these customers will ultimately generate economic returns on these investments.
A critical factor is depreciation policy. Tech giants have extended server lifespans to six years on their books. However, Nvidia's product cycles now operate on an annual basis, potentially rendering older chips functionally obsolete and less energy-efficient long before they are fully written down, according to Scion.
Nvidia has countered this claim, asserting that its hardware remains productive for a longer duration due to efficiencies driven by its CUDA software system.
Burry and other critics point to an apparent contradiction: Nvidia promotes the superior performance and efficiency of its latest chips while simultaneously maintaining that older chips remain economically viable. They argue that one of these claims must be inaccurate.
Burry has launched a new Substack newsletter to expound upon his bearish perspective on AI. While the long-term fate of generative AI remains uncertain, Burry is once again positioning himself on the cautious side of a rapidly evolving narrative.
