AI-Driven Overbuilding and GPU Depreciation Risks: A Cautionary Tale from Michael Burry

The current AI boom, fueled by unprecedented investment in hyperscale infrastructure, has drawn comparisons to historical financial bubbles. Michael Burry, the investor who famously predicted the 2008 housing crisis, has sounded the alarm, warning that the sector's rapid expansion and aggressive capital expenditures mirror the speculative excesses of the dot-com era and the pre-2008 housing market. His concerns center on GPU depreciation risks, accounting practices, and systemic fragility-issues he argues could lead to a correction akin to past crashes.

Parallels to the 2008 Crisis

Burry's analogy to the 2008 crisis hinges on the interplay between speculative behavior and flawed accounting. In his Substack newsletter Cassandra Unchained, he highlights how tech giants have extended the useful life of AI-related GPU assets, artificially inflating earnings by spreading depreciation costs over longer periods. For instance, Meta extended its AI chip depreciation period from 4.5 to 5.5 years, saving $2.9 billion in 2025. Similarly, Google and Microsoft now depreciate their GPUs over six years, a stark contrast to the real-world obsolescence cycle, where Nvidia releases new chip generations annually according to analysis.

This practice, Burry argues, echoes the pre-2008 housing market, where overleveraged institutions masked fragility through optimistic accounting assumptions. "The useful life of these assets is likely only 2–3 years," he warns, estimating a potential $205 billion cumulative write-down if reality aligns with his thesis. Such a scenario could trigger a cascade of earnings downgrades and investor panic, much like the mortgage-backed securities collapse in 2008.

GPU Depreciation: A Hidden Time Bomb

The core of Burry's critique lies in the mismatch between accounting assumptions and technological reality. According to Bloomberg, the rapid innovation cycle in AI hardware means older GPUs lose value far faster than their reported depreciation schedules suggest. This discrepancy creates a "hidden risk" for tech firms, as their reported profits may be artificially inflated while actual asset values erode according to market analysis.

Barclays analysts have estimated that GPU depreciation costs could reduce earnings for major tech companies by up to 10%. Burry's calculations are even more dire: he projects a $176 billion understatement of depreciation costs by 2028 if current accounting practices persist. This risk is compounded by the sector's reliance on multi-year vendor contracts and continuous reinvestment, which could strain cash flows if GPU obsolescence accelerates.

Counterarguments: Real Revenue vs. Speculative Hype

Not all observers share Burry's pessimism. Forbes argues that unlike the dot-com bubble, many AI companies are already generating substantial revenue through multi-year contracts, providing a buffer against speculative downturns. Additionally, proponents of the AI boom emphasize that infrastructure investments are driven by a "fundamental shift in energy-based computation," not just hype according to industry analysis.

Nvidia, a key target of Burry's bearish bets, has denied allegations of artificially inflating demand through vendor financing. The company's recent Blackwell GPU launch, which promises significant performance gains, has been hailed as a catalyst for sustained growth. However, Burry's analogy of Nvidia to Cisco during the dot-com era underscores his belief that current valuations are built on unsustainable assumptions.

Implications for Investors

For investors, the debate underscores the need for caution. While AI's transformative potential is undeniable, the sector's financial health depends on aligning accounting practices with technological realities. Burry's $1.1 billion bet against AI-focused companies reflects his conviction that a correction is inevitable. However, the counterarguments highlight the importance of distinguishing between speculative overbuilding and genuine innovation.

Policymakers and investors alike must grapple with the question: Are we witnessing a bubble, or a paradigm shift? The answer may lie in how quickly the market adjusts to the realities of GPU depreciation and competitive pressures. As Burry's warnings suggest, the parallels to 2008 are not coincidental-they are a reminder that history often repeats itself when fundamentals are obscured by optimism.