Michael Burry's Big Tech Accounting Criticism: AI Depreciation Schedule Analysis

This analysis is based on recent reports from Morningstar/MarketWatch [1], Fortune [2], and InvestorPlace [3] published in November 2025, examining Michael Burry’s allegations regarding Big Tech’s accounting practices for AI infrastructure.

Michael Burry, renowned for predicting the 2008 financial crisis, has identified what he characterizes as “one of the more common frauds of the modern era” [1]. His analysis focuses on major technology companies—Meta (META), Alphabet (GOOGL), Microsoft (MSFT), Amazon (AMZN), and Oracle (ORCL)—extending depreciation schedules for AI servers and GPUs from traditional 2-3 year periods to 5-6 years [1][2][3].

• Burry projects this accounting change will understate depreciation by $176 billion between 2026-2028 [2][3]
• Potential earnings inflation: Oracle by 27%, Meta by 21% [2][3]
• Current market data shows these companies trading at elevated multiples: Meta P/E 26.98x, Microsoft P/E 36.23x, Alphabet P/E 27.26x, Amazon P/E 33.58x [0]

Despite Burry’s warnings, market reactions have been mixed. As of November 14, 2025:

• NASDAQ gained +1.58%, suggesting the market hasn’t fully embraced the bearish thesis [0]
• Individual tech stocks showed weakness: META down 15.06% over the past month, AMZN -1.22%, GOOGL -0.78%, while MSFT showed relative strength at +1.37% [0]

• Jim Morrow of Callodine Capital describes this as “the most crowded trade in history” with a coming “tsunami of depreciation” [2]
• Richard Jarc of Uncovered Alpha estimates GPU economic life at 1-2 years, not the 5-6 years being booked [2]
• The Economist labeled this “the $4 trillion accounting puzzle at the heart of the AI cloud” [2]

• Bank of America’s semiconductor team maintains AI demand remains “robust” with Nvidia disclosing $500 billion+ in 2025-26 data-center orders [2]
• CoreWeave CEO reports strong demand for older chips, with H100s from 2022 still commanding premium prices [1]
• Some argue unprecedented AI demand means even older chips retain value for less demanding tasks [2]

The core issue centers on the mismatch between accounting assumptions and technological obsolescence. Documented changes include:

• Meta extended server/network asset useful lives from 4-5 years to 5.5 years effective January 2025 [2]
• Google extended from 3 to 6 years [2]
• Oracle extended from 5 to 6 years [2]

This creates a fundamental tension: while accounting rules may permit longer depreciation schedules, the rapid pace of AI advancement suggests technological obsolescence occurs much faster. Meta’s depreciation expense progression—$6.39 billion in 2020, $7.56 billion in 2021, $8.50 billion in 2022 [2]—shows the scale of these charges before the extensions.

The concentration risk is particularly concerning, with nearly half of 401(k) money tied to six megacaps [2]. The unprecedented capital deployment—hyperscale data centers costing $50 billion each, with companies attempting “fifty of them at once” [2]—creates systemic exposure if the AI investment thesis proves flawed.

1. Accounting Risk
   : If SEC or accounting standard bodies rule current depreciation practices inappropriate, companies could face massive restatements and accelerated depreciation charges.

2. Technology Obsolescence Risk
   : If AI hardware advances faster than expected, companies may need to write down assets or accelerate depreciation, significantly impacting earnings.

3. Concentration Risk
   : Extreme market concentration in Big Tech creates systemic vulnerability if these companies face simultaneous challenges.

4. Regulatory Scrutiny
   : While no current SEC investigations have been announced, the high-profile nature of these allegations could trigger regulatory review.

• SEC statements or investigations into tech accounting practices
• Company earnings calls addressing depreciation methodology
• Technical analysis of AI hardware longevity from independent experts

• Actual refresh cycle timing vs. projected depreciation schedules
• Free cash flow conversion rates vs. reported earnings
• Institutional investor positioning changes

• Power grid capacity and data center utilization rates
• AI ROI metrics and actual productivity gains
• Potential government intervention or support mechanisms

The debate over Big Tech’s accounting practices highlights fundamental questions about earnings quality in the AI era. While companies argue that extended depreciation schedules reflect actual useful life in less demanding applications, critics contend this masks the true cost of rapid technological obsolescence.

The coming 2026-2027 period will be crucial as companies face decisions about refreshing AI infrastructure purchased during the recent boom. If Burry’s analysis proves correct, investors could face significant corrections as accounting reality catches up with technological obsolescence.

Unlike previous tech bubbles, the AI sector has unprecedented government support and strategic importance, which could provide a backstop against normal market corrections [2]. However, the fundamental mismatch between accounting assumptions and technological reality suggests investors should scrutinize earnings quality metrics, particularly depreciation trends and free cash flow versus reported earnings.