Michael Burry, the enigmatic investor immortalized for his contrarian bet against the 2008 housing market, has once again captured the financial world's attention, this time by taking a substantial short position on the artificial intelligence industry. This move, highlighted by Matthew Berman in a recent video, immediately raises a critical question for founders, venture capitalists, and tech professionals: Is the AI boom merely a speculative bubble, or does it represent a foundational shift, albeit with some inevitable market froth? Berman’s commentary deftly navigates this complex terrain, drawing parallels to historical economic cycles and dissecting the underlying fundamentals of AI’s rapid ascent.
The discussion begins by establishing a clear definition of an economic bubble, courtesy of ChatGPT: "An economic bubble is when the price of an asset... rises far above its real or sustainable value because of excessive demand, speculation, and hype." Berman then outlines the typical stages: displacement (new innovation), boom (prices rise rapidly), euphoria (speculation and greed dominate, prices detach from fundamentals), profit-taking (smart investors sell), and panic (prices collapse). He differentiates between two types of bubbles: those, like the 1929 stock market crash, that lack fundamental infrastructure build-out and cause irreparable damage, and those, such as the dot-com bubble, which were ultimately mis-timed infrastructure plays that eventually bore fruit.
In assessing the current AI landscape, Berman points to an undeniable surge in infrastructure investment. Bank of America, for instance, "now sees global hyperscale spending rising 67% in 2025 and another 31% in 2026, with total outlays climbing to $611 billion." This monumental capital allocation is directed towards data centers, power generation, and, crucially, advanced chips. Unlike the 1929 scenario, the AI boom is underpinned by tangible infrastructure development, a critical distinction that suggests a more robust, albeit potentially overvalued, foundation.
However, the question of whether prices have detached from fundamentals remains central. Consumer adoption of AI tools like ChatGPT has been nothing short of explosive, with OpenAI planning to hit one billion users by the end of 2025. A Menlo Ventures report further substantiates this, revealing that "more than half of American adults (61%) have used AI in the past six months, and nearly one in five rely on it every day." This level of immediate, widespread engagement suggests a genuine demand for AI’s utility. Enterprise adoption, while slower due to the complexities of integration and security, is also steadily advancing, indicating that the value proposition is being recognized across various sectors.
A deeper look into the intricate financial flows within the AI ecosystem reveals what Berman dubs an "AI money machine." An infographic from Bloomberg illustrates how key players like Nvidia, Microsoft, and OpenAI are deeply intertwined through investments, services, and hardware purchases. Nvidia, as the dominant chip supplier, sits at the center, selling its GPUs to data center companies, cloud providers, and AI model developers. What’s intriguing—and potentially concerning—is the pattern of investments where Nvidia invests in AI startups, which then turn around and purchase Nvidia chips and services. This circular flow raises questions about whether true new value is being created at every step, or if capital is largely recirculating within a closed system, inflating valuations.
Michael Burry's specific critique focuses on accounting practices, arguing that AI hyperscalers are "understating depreciation by extending useful life of assets artificially boosts earnings—one of the more common frauds of the modern era." His contention is that the rapid product cycles of Nvidia chips and servers should not justify extended useful lives for depreciation purposes, suggesting an artificial inflation of earnings. Berman counters this by highlighting that many older generation chips are still running at 100% utilization, and that the sheer demand for compute power means even "last-generation chips" are being fully utilized. This points to a genuine need for processing power, rather than just speculative overstatement.
Adding another layer of complexity is the emerging constraint of power. "Microsoft CEO Satya Nadella said... the problem in the AI industry is not an excess supply of compute, but rather a lack of power to accommodate all those GPUs." This highlights a very real, physical bottleneck in scaling AI infrastructure, transcending mere financial speculation. The "Magnificent Seven" tech giants (Meta, Microsoft, Nvidia, Apple, Tesla, Amazon, Alphabet) are at the forefront of this AI-driven economic expansion, their collective growth heavily influencing the US GDP. This concentration of market power, while impressive, also warrants caution.
Ultimately, Berman concludes that while aspects of the AI boom might exhibit speculative characteristics, it fundamentally differs from a purely destructive bubble like 1929. The massive infrastructure build-out, coupled with genuine and growing demand from both consumers and enterprises, aligns it more with the internet boom of the late 90s—a period of over-exuberance that ultimately laid the groundwork for a truly transformative technology. The core insight is that AI, described as "the most important technology humans have ever created," offers too much tangible opportunity for its current trajectory to be dismissed as mere hype. The challenges, such as power constraints and the long-term sustainability of interconnected investments, are real, but they are problems of scaling a valuable innovation, not signs of its inherent worthlessness.
