Michael Burry, famed for predicting the 2008 financial crisis, has once again targeted tech giants. On November 11, Burry posted on social media that these companies are artificially inflating earnings by underestimating depreciation through extending the "useful life" of assets. He estimates that from 2026 to 2028, major tech firms could overstate profits by $176 billion due to this practice.
Burry promised to reveal more details on November 25. He specifically highlighted that by 2028, Oracle's profits might be overstated by 26.9%, while Meta's could be exaggerated by 20.8%. Notably, Burry previously disclosed holding put options on NVIDIA and Palantir, reigniting market attention on his bearish stance toward the AI sector.
Wall Street institutions like Morgan Stanley and Bank of America have also warned that the market severely underestimates the true scale of current AI investments and is unprepared for the impact of future depreciation expenses. This could lead to tech giants' actual profitability falling far below market expectations.
**Accounting "Tricks": Extending Hardware Lifespans to Smooth Costs** Burry argues that tech giants are employing accounting "gimmicks." In his post, he noted that extending the "useful life" of assets to understate depreciation is "one of the most common modern frauds." He emphasized that computing equipment, such as NVIDIA chips and servers—typically with 2-3 year lifecycles—should not have prolonged depreciation periods. However, "hyperscalers" like Meta, Alphabet, Microsoft, Oracle, and Amazon are doing just that, with some even stretching depreciation cycles to six years.
Amazon, however, shortened the estimated useful life of some servers and networking equipment from six to five years in Q1 2025, citing accelerated AI advancements. If this trend reverses and expands, it could lead to accelerated depreciation recognition, impacting short-term earnings.
**Depreciation "Time Bomb": Market Unprepared for Future Profit Shock** Bank of America has also warned that Wall Street is "underestimating" the growth of future depreciation expenses. Analyst Justin Post noted that as Google, Meta, and Amazon's capital expenditures surge in 2024 and 2025, their depreciation costs will inevitably accelerate post-2026. By 2027, the market's consensus for these three companies alone could understate actual depreciation expenses by nearly $16.4 billion, suggesting their true profitability may lag far behind current expectations.
The "short lifespan" of AI assets exacerbates this risk. Hardware like GPUs, used for AI computing, faces rapid technological obsolescence and high workloads, with effective lifespans of just 3-5 years—contrary to tech giants' extended depreciation practices.
**Wall Street's Warning: Underestimated Capex and "Off-Balance-Sheet" Expansion** Morgan Stanley's research highlights that tech giants' AI capital expenditure intensity is nearing peak levels seen during the dot-com bubble, yet public data doesn't fully reflect the investment picture. Two key factors contribute to this underestimation:
1. **Rise of Financing Leases**: Companies like Microsoft and Oracle are increasingly using these "off-balance-sheet" tools to build data centers, with initial investments excluded from traditional capex. Including financing leases, Morgan Stanley estimates Microsoft's 2026 capex-to-sales ratio would jump from 28% to 38%. 2. **Delayed Impact of "Construction in Progress (CIP)"**: Significant capital remains on balance sheets without depreciation, meaning the profit impact has only just begun.
Bank of America warns that the AI infrastructure market could repeat historical patterns of overcapacity and price wars due to aggressive investment. If supply continues outpacing demand, aggressive pricing strategies may emerge by 2027, forcing hyperscalers to cut prices to maintain data center utilization—eroding profitability. This further supports Burry's argument about overstated tech earnings.