The investment community has been set ablaze by a series of observations from Michael Burry, the famed investor who successfully predicted the 2008 housing market collapse. Burry is now casting a shadow of doubt over the current artificial intelligence boom, specifically targeting the astronomical rise of Nvidia. His analysis draws a direct line between the current semiconductor frenzy and the infrastructure build-out that defined the late 1990s, suggesting that history may be preparing to repeat itself in a painful fashion.
At the heart of Burry’s thesis is the concept of ‘purchase commitments.’ During the height of the dot-com bubble, Cisco Systems was the undisputed king of the networking world. As telecommunications companies and early internet startups scrambled to build the digital highway, Cisco’s hardware became the essential component of the era. To ensure they wouldn’t be left behind, companies placed massive advance orders, creating a backlog that suggested infinite growth. Burry notes that Nvidia is currently experiencing a nearly identical phenomenon as tech giants and sovereign nations race to secure H100 and Blackwell chips.
The danger, according to this perspective, lies in the artificial inflation of demand. When supply is tight, customers often over-order or double-order to ensure they receive at least some allocation of the necessary hardware. This creates a feedback loop where the manufacturer sees a massive spike in future revenue commitments, which in turn drives the stock price to levels that assume these growth rates are permanent. In the case of Cisco, when the initial infrastructure was finally completed, the ‘purchase commitments’ vanished almost overnight, leading to a catastrophic inventory glut and a share price collapse from which it took decades to recover. 
Nvidia currently finds itself in a unique position where its primary customers are also its potential competitors. Companies like Microsoft, Amazon, and Google are currently Nvidia’s largest buyers, yet they are simultaneously developing their own custom AI silicon to reduce dependency on the Santa Clara-based chipmaker. Burry’s comparison suggests that we are currently in the ‘build phase’ of the AI revolution. Once the major data centers are populated with enough GPUs to handle the initial wave of large language models, the frantic pace of purchasing is likely to normalize or drop off entirely.
Market analysts have pointed out that Nvidia’s margins are currently at historic highs, a feat rarely sustainable in the hardware sector. Burry’s skepticism focuses on the fact that while the technology is undoubtedly transformative, the financial architecture supporting the stock price is built on the assumption that the current buying spree is a baseline rather than a peak. If the massive investments in AI do not yield immediate and substantial returns for the software companies buying the chips, the capital expenditure budgets will eventually be slashed.
While many bulls argue that AI is a fundamentally different beast than the fiber-optic build-out of 1999, the structural similarities in market behavior are difficult to ignore. The concentration of market gains in a single hardware provider mirrors the dominance Cisco once held. In both instances, the narrative of a ‘new era’ was used to justify valuations that ignored traditional cyclical patterns in the semiconductor and networking industries.
Michael Burry has never been one to shy away from contrarian positions, often enduring periods of significant underperformance while waiting for his structural theses to play out. Whether his comparison to the dot-com era is premature or prophetic remains to be seen, but his focus on purchase commitments serves as a sobering reminder that today’s backlogs can quickly become tomorrow’s surplus. For investors heavily weighted in the semiconductor space, the ghost of Cisco’s past offers a cautionary tale about the volatility of infrastructure cycles.
