Michael Burry's Big Tech Accounting Critique: AI Investment Quality Concerns

This analysis is based on recent market reports [1][2] published in November 2025, examining Michael Burry’s critique of Big Tech accounting practices and their implications for AI investment quality.

Michael Burry, renowned for predicting the 2008 housing crisis, has emerged from a two-year silence to challenge the earnings quality of major technology companies [1][2]. His critique centers on systematic accounting changes across Meta, Alphabet, Amazon, and Microsoft that have extended depreciation schedules for servers and GPUs, thereby reducing non-cash charges and artificially boosting reported net income [1][2].

The timing of Burry’s intervention is particularly significant given the unprecedented AI infrastructure spending boom. The four largest AI infrastructure spenders are projected to increase combined capital expenditures by approximately 40% to $460 billion in the next 12 months [1]. This massive investment surge coincides with depreciation expenses that have already doubled from about $10 billion in Q4 2023 to nearly $22 billion in the most recent quarter, with analysts projecting almost $30 billion by this time next year [1].

A fundamental disconnect exists between accelerated technological obsolescence and extended accounting timelines. Nvidia now releases new chips every 12-18 months instead of every 2 years, suggesting hardware becomes obsolete faster, not slower [2]. This creates a potential earnings cliff where companies may face massive write-downs as technological reality outpaces accounting treatment.

Not all tech companies have followed the same path:

• Meta
  : Extended server/network asset useful life from 4-5 years to 5.5 years in January 2025, estimating this would reduce 2025 depreciation expense by $2.9 billion [1][2]
• Microsoft
  : Extended server depreciation from 4 to 6 years in 2022 [2]
• Alphabet
  : Made similar extensions in 2023 [2]
• Amazon
  : Actually shortened useful life to 5 years from 6 years in February 2025, taking a more conservative approach [1]

The magnitude of these investments presents unprecedented operational challenges. No company has ever managed a $50 billion project before, and these firms are attempting fifty such projects simultaneously [2]. Additionally, data centers in Santa Clara and Northern Virginia are sitting idle waiting for grid hookups that could take years, meaning billions in GPU investments aren’t generating returns [2].

1. Earnings Quality Risk
   : Burry estimates that Big Tech will understate depreciation by $176 billion between 2026-2028, with specific company impacts including Oracle (earnings overstated by 26.9%) and Meta (earnings overstated by 20.8%) [2].

2. Concentration Risk
   : Nearly half of all 401(k) money is effectively tied to six megacaps, making this “the most crowded trade in history” according to some analysts [2].

3. Infrastructure Bottleneck Risk
   : Every month a $35 billion GPU stack sits without power represents approximately $1 billion in depreciation “burning a hole” on balance sheets [2].

1. Conservative Accounting Advantage
   : Amazon’s decision to shorten depreciation schedules may provide more transparent earnings visibility compared to peers [1].

2. Market Correction Potential
   : The Economist estimates that if these assets were depreciated over 3 years instead of extended timelines, annual pre-tax profits would fall by $26 billion (roughly 8% hit), potentially creating entry points for disciplined investors [2].

3. Infrastructure Development
   : Companies that successfully solve power grid connectivity challenges may gain significant competitive advantages in AI infrastructure deployment [2].

The debate over Big Tech earnings quality centers on whether current accounting practices reflect economic reality or mask underlying investment inefficiencies. While Mag7 earnings are still projected to grow ~27% YoY despite massive AI capex and rising depreciation costs, questions persist about profit sustainability [1][2].

Market reaction has been mixed, with Meta significantly underperforming (up only 3% year-to-date versus Nasdaq 100’s 18% gain), while Alphabet has performed relatively better (up 46% year-to-date) [1]. The broader market has shown resilience, with the S&P 500 up 0.93% and Nasdaq up 1.58% on November 14, suggesting that Burry’s concerns haven’t yet triggered broader market panic [0].

Historical precedents from previous capital spending booms in shale, fiber, and railroads suggest that current AI infrastructure investment patterns may follow similar trajectories of “overcapacity, low returns, and bailouts” [2]. However, optimists argue that current spending is stimulative for future earnings and that the trade is less crowded than critics claim [1][2].

The fundamental question remains: Are current AI investments generating genuine productivity gains, or merely masking capital intensity through accounting choices? The answer to this question will likely determine whether Burry’s warnings prove prescient or whether the current AI investment cycle delivers on its promised returns.