Michael Burry estimates major US hyperscalers will understate depreciation expenses by approximately $176 billion between 2026-2028 by extending useful life assumptions for AI compute infrastructure according to his analysis of SEC filings posted November 2025 on @michaeljburry. The practice inflates reported earnings without improving underlying cash generation, creating accounting-driven growth that markets treat as operational performance.

The warning centers on balance sheet mechanics rather than technology skepticism. AI infrastructure investments accelerated dramatically through 2024-2025, with companies purchasing Nvidia GPU servers on 2-3 year replacement cycles. However, depreciation schedules extended from historical 3-4 year useful lives to 5-6 years, spreading costs across longer periods and boosting near-term earnings artificially.

How Extended Depreciation Inflates Earnings

Depreciation represents non-cash expense reflecting asset value decline over time. When companies extend depreciation schedules, annual expense decreases while reported operating income rises. Cash flows remain unchanged since depreciation doesn't affect actual spending, but GAAP earnings improve mechanically through accounting assumption changes.

According to Burry's analysis of company filings, Meta extended network and compute equipment useful life from 3-4 years in 2020-2021 to 4.5-5.5 years by 2024-2025. Oracle increased depreciation schedules from 5 years to 6 years. Microsoft extended equipment life from 3-4 years to 6 years. Google maintained 6-year schedules while increasing capital intensity. Amazon reduced depreciation period from 6 years to 5 in 2024 but remains above historical 4-year standard.

The table Burry shared shows systematic extension across major cloud and AI infrastructure operators. This timing coincides with massive AI capex ramp, suggesting companies are managing earnings optics during period of unprecedented infrastructure spending.

When AI servers costing $250,000-500,000 each depreciate over 6 years instead of 3, annual expense drops 50%. Multiply across thousands of units and billions in annual capex, the earnings impact becomes material. Burry's $176 billion estimate through 2028 reflects accumulated difference between extended schedules and historical practices.

Why This Creates 2026-2028 Earnings Risk

The accounting choice delays expense recognition but doesn't eliminate it. If AI equipment actually requires replacement after 3-4 years due to obsolescence, technological advancement, or performance degradation, companies will face write-downs when extended useful life assumptions prove incorrect.

Burry estimates Oracle will overstate earnings 26.9% by 2028 under current depreciation assumptions. Meta faces approximately 20.8% overstatement risk. These aren't fraud allegations but warnings about optimistic accounting that markets haven't priced into valuations.

The dot-com comparison emerges here. During 1999-2000, telecom and infrastructure companies capitalized spending aggressively and extended depreciation to show profitability. When equipment values collapsed faster than assumed, write-downs triggered across sector. Burry sees similar pattern: rapid technological change making long depreciation schedules unrealistic.

NVIDIA releases new GPU architectures every 18-24 months. Each generation offers substantial performance improvements, making previous generation less competitive for AI workloads. Companies betting 6-year useful lives assume either technological progress will slow dramatically or that 3-4 year old hardware will remain economically viable despite newer, more efficient alternatives. History suggests both assumptions are optimistic.

Market Structure Amplifies Valuation Risk

Burry's concern extends beyond individual company accounting to market structure dynamics. According to his analysis, extreme concentration in passive investment vehicles creates fragility that didn't exist during previous bubbles.

In 2000, speculation distributed across hundreds of internet stocks. Today, AI exposure concentrates in 5-7 mega-cap names that dominate major indices. When investors want AI exposure, they buy MSFT, GOOGL, META, NVDA, ORCL regardless of valuation. Passive funds allocate mechanically based on market cap weighting, creating self-reinforcing inflows during uptrends.

This structure works during calm markets but becomes unstable during volatility. Passive vehicles don't evaluate fundamental value—they track indices. When redemptions begin, selling occurs automatically across all holdings. The same concentration that amplified gains during rally magnifies losses during correction.

Burry argues this makes current setup potentially more dangerous than 2000 despite stronger underlying businesses. Tech companies today generate real revenue and cash flow unlike many dot-com era firms. But concentration means systematic risk sits in fewer names, creating single points of failure for portfolio performance.

Burry's Positioning Shows Conviction Beyond Rhetoric

According to Scion Asset Management 13F filings, Burry holds substantial put option positions on broad market indices and specific large-cap technology stocks. These aren't small hedges but meaningful bearish bets indicating expectations of volatility markets aren't pricing.

Put options cost money and decay over time, making them expensive to maintain. Burry's willingness to pay these costs signals conviction about structural imbalance rather than short-term market timing. His positioning suggests he expects not just correction but potential disorderly repricing when accounting realities emerge.

The timing element matters. Burry stated additional detail would come November 25th in his original post. This suggests specific catalysts or analysis he planned to release publicly, potentially including detailed company-by-company depreciation analysis or earnings adjustment models.

Implications for Crypto Markets

Crypto's correlation with risk assets means AI bubble concerns translate directly into digital asset volatility expectations. Historical data from previous equity corrections shows crypto typically amplifies downside during liquidity stress events.

During March 2020, May 2021, and June 2022 corrections, Bitcoin declined 30-50% while tech stocks fell 15-25%. Altcoins experienced 60-80% drawdowns. This pattern reflects crypto's position as leveraged expression of risk appetite rather than true alternative asset during stress periods.

If Burry's depreciation analysis proves correct and earnings revisions hit AI stocks during 2026-2028, crypto would likely face three specific pressures. Forced deleveraging across perpetual futures markets would trigger liquidation cascades. Institutional allocators reducing overall risk exposure would sell crypto alongside equities. Retail sentiment would sour as losses mount across portfolios.

The counterargument is that Bitcoin specifically might benefit as alternative to equity risk during late-stage repricing. However, this assumes crypto reaches sufficient maturity to decouple from traditional risk assets. Current market structure suggests decoupling remains aspirational rather than actual.

The Counterargument About Real Business Quality

The legitimate bull case acknowledges Burry's points while arguing fundamentals justify valuations. Unlike 2000, today's tech leaders operate profitable businesses at scale. Meta generates $40+ billion annual free cash flow. Microsoft and Google produce similar amounts. Apple sits on $150+ billion net cash. These aren't speculative ventures burning through runway.

AI adoption may also deliver genuine productivity improvements that justify current investment levels. If AI increases corporate efficiency 10-20% across economy, current spending represents reasonable bet on future returns. Historical precedent from computing, internet, and mobile revolutions shows technology skepticism often proves too conservative.

Burry doesn't dispute strong underlying businesses. His argument is narrower: markets may already price multiple waves of future growth while ignoring how much reported earnings depend on optimistic accounting assumptions. Strong business can coincide with overextended valuation, especially when accounting choices mask true expense recognition timing.

Why Questioning Depreciation Matters

Most investors ignore depreciation as accounting detail rather than valuation signal. Burry's focus on this specific line item reflects his approach: find mispriced risk in places consensus overlooks. Extended depreciation schedules are public information in SEC filings, yet few analysts adjust earnings models to reflect realistic equipment replacement cycles.

This creates information asymmetry favoring those who dig into footnotes. When write-downs eventually occur, markets will react as if surprised despite public availability of underlying assumptions. The pattern repeats: accounting flexibility smooths earnings during good times, then concentrates pain during corrections.

Burry's timing may prove early—his warnings often precede market recognition by months or years. But the imbalance he identifies is documented in company filings rather than speculative. When dominant narrative meets heavy passive flows meets accounting assumptions smoothing earnings meets elevated confidence, historical outcomes skew negative.

Whether AI boom justifies valuations or sets up repricing depends largely on depreciation assumptions proving realistic. Burry bets they won't. Markets priced perfection into AI trade while ignoring $176 billion in potential earnings revisions ahead. One side will be proven right through 2026-2028. The cost of being wrong on either side is substantial.