NVIDIA Depreciation Accounting Concerns: Michael Burry's AI Chip Profitability Warning

This analysis examines Michael Burry’s recent warning about AI chip depreciation practices and their potential impact on NVIDIA’s business model. The core issue centers on a fundamental accounting mismatch between NVIDIA’s rapid innovation cycle and customers’ depreciation assumptions [1][2].

NVIDIA operates on a 1-year chip production cycle with AI chips having an actual useful life of 2-3 years due to rapid technological obsolescence [1][3]. However, major hyperscalers including Meta, Google, Oracle, Microsoft, and Amazon are depreciating this equipment over 5-6 years, effectively understating annual depreciation expenses by approximately 50% [3]. This creates a significant “accounting subsidy” that inflates near-term profitability [3].

Google engineers estimate GPUs running at standard AI workload utilization (60-70%) survive only 1-2 years, with 3 years as the maximum due to thermal and electrical stress [3]. NVIDIA’s Blackwell chips provide 4-5x faster inference than the previous H100 generation, making older hardware economically obsolete even if still functional [3].

Burry estimates major hyperscalers will understate depreciation by approximately $176 billion between 2026-2028 [1]. Specific projections include Oracle potentially overstating earnings by 26% and Meta by 20% due to this practice [1]. Barclays analysts have already cut earnings forecasts for AI firms by up to 10% for 2025 to account for more realistic depreciation assumptions [3].

The depreciation-useful life mismatch represents more than just an accounting issue—it could trigger systemic financial consequences. If AI capex write-offs occur when financed with debt, it could trigger defaults and broader market disruptions [2]. The scale is substantial, with AI infrastructure spending estimated at $300+ billion in 2025 alone, with computing hardware representing 60% of these costs [3].

NVIDIA’s $110 billion in disclosed investments and financing commitments represent 67% of revenue, creating potential circular financing risks if customers face cash constraints [3]. This vendor financing model could amplify the impact when accounting adjustments occur.

Bain research estimates AI firms face an $800 billion annual revenue gap by 2030 to fund capital expenses, growing to over $1.5 trillion if true costs are properly accounted for [3]. This suggests the current business model may be fundamentally unsustainable without significant revenue growth or cost reductions.

• Earnings Shock Risk:
  When accounting catches up to operational reality, hyperscalers could face major earnings hits that cascade to NVIDIA’s revenue stream [1][2]
• Customer Financial Health:
  The impact on hyperscalers’ actual cash flows and ability to continue AI spending represents a significant demand risk for NVIDIA [1]
• Regulatory Intervention:
  Potential accounting standard changes or regulatory investigations could force accelerated depreciation across the industry [3]

• NVIDIA’s Q3 2026 earnings report on November 19, 2025, will be crucial for assessing demand sustainability [4]
• Changes in depreciation assumptions or capex guidance from major hyperscaler customers
• Real-world chip replacement rates and performance degradation data from actual deployments

Big Tech argues that 5-6 year useful lives are reasonable because GPUs serve multiple purposes beyond AI training, and older chips can be repurposed from training to inference workloads [3]. Some analysts suggest this may represent legitimate accounting differences rather than intentional manipulation [3].

The depreciation-useful life mismatch represents a significant structural risk that could materially impact NVIDIA’s revenue stream if hyperscaler customers face cash constraints or are forced to accelerate depreciation [1][2][3]. The $176 billion estimated understatement of depreciation through 2028 suggests this is not a minor accounting issue but could represent a fundamental mispricing of risk across the AI ecosystem [1].

NVIDIA is currently trading at $189.20 with strong analyst consensus (75.9% “Buy” ratings) and an average price target of $235.00 [0]. However, the depreciation concerns could impact this bullish sentiment if customers face earnings adjustments that constrain their AI spending capabilities.

The situation requires careful monitoring of upcoming earnings reports, hyperscaler guidance changes, and potential regulatory developments that could accelerate the recognition of these hidden costs [1][2][3].