Michael Burry's AI Bubble Warning: The $176 Billion Depreciation Controversy

on

 

Michael Burry's AI Bubble Warning: The $176 Billion Depreciation Controversy

On November 10, 2025, I paused while watching a YouTube video about Michael Burry's latest post. "Artificially extending the useful life of assets to understate depreciation is one of the common scams of the modern era," he declared, bringing back memories of his sharp prediction of the 2008 subprime mortgage collapse. This time, the AI industry was his target. He claimed that tech giants like Amazon, Microsoft, Google, Meta, and Oracle were inflating their profits through accounting tricks while pouring hundreds of billions into AI infrastructure.

But I don't take Burry's words at face value. Since his subprime success, he's made numerous similar warnings that didn't pan out. He's not a prophet—he's an investor. More specifically, a short seller who profits most when his positions prove correct. His warnings are grounded in facts, but they're also carefully crafted messages designed to serve his own interests. In this essay, I'll dissect Burry's claims, examine whether an AI bubble truly exists, and explore what we as investors should believe—all through my own analytical lens. (This is purely my personal analysis and not investment advice.)

The Man Behind the Curtain: Prophet or Strategist?

To understand Michael Burry, you need to look at his track record. From 2005 to 2007, he bet against the subprime mortgage market by purchasing credit default swaps (CDS). Wall Street mocked him. But when the 2008 crisis hit, his fund Scion Capital posted a 489% return. He's the character Christian Bale portrayed in "The Big Short." Since then, he's earned the nickname "Cassandra"—the tragic Greek prophet who could see the future but was never believed.

But I don't see Burry as Cassandra. He's a strategist. In 2021, before inflation hit, he shorted bonds. In 2022, deeming Tesla overvalued, he bought put options. And now in 2025, he holds put options on Nvidia and Palantir. See the pattern? Burry bets on downturns he's convinced of, then publicly announces his conviction. His X posts aren't just warnings—they're marketing for his positions.

This isn't necessarily bad. Investors act on their analysis and bear responsibility for the results. Burry does the same. But when we listen to him, we must always consider his position. Even if his warnings are 70-80% factual, the remaining 20-30% might be embellishment designed to maximize his own profits.

The $176 Billion Depreciation Gap: Burry's Math

Burry's core argument is clear. The AI boom has hyperscalers like Amazon, Microsoft, Google, Meta, and Oracle pouring massive capital into Nvidia GPUs and servers. The problem? They're artificially extending the "useful life" of this hardware in their accounting, thereby reducing annual depreciation expenses. The result: inflated net income.

Let's get specific. AI hardware, particularly GPUs, has an actual product cycle of 2-3 years. Nvidia itself suggests about 3 years as the economic lifespan for datacenter GPUs. The reason is simple: AI model size and complexity doubles annually. If you could train 12 large language models with 1,000 H100 GPUs in 2023, you'd need 4,000-5,000 for the same work in 2025. A 2-year-old GPU loses 75-80% of its economic value.

Yet these companies set their accounting depreciation schedules at 4-6 years:

  • Meta: 5 years
  • Microsoft: 5-6 years
  • Google: 4-6 years
  • Amazon: 5 years
  • Oracle: 6 years

Burry calculates this will result in $176 billion in understated depreciation between 2026-2028. The formula is straightforward: Capital expenditure × (1/actual lifespan - 1/accounting lifespan). For example, if you depreciate a $1 billion asset over 6 years instead of 3, annual expenses drop from $333 million to $167 million. That $166 million difference gets added to net income.

The profit overstatement by company:

  • Oracle: 26.9% by 2028
  • Meta: 20.8%
  • Amazon, MS, Google: 15-25%

This calculation is transparent and logical. Burry analyzed capital expenditures and depreciation policies based on each company's 10-K filings. Numbers don't lie.

But Is It Really "Fraud"?

Burry calls this "one of the common scams of the modern era." I disagree with that characterization. US GAAP (Generally Accepted Accounting Principles) allows companies to "reasonably estimate" asset useful lives. These companies disclose changes in footnotes to their 10-K reports and pass audits. It's not illegal.

So what's their defense? Companies argue that "GPUs can be redeployed from training to inference for 4-5 more years of use." Indeed, services like Google Colab's free tier and OpenAI's GPT-3.5 utilize older GPUs. While unsuitable for training cutting-edge models, they're sufficient for running inference on existing models.

Additionally, with GPU demand exceeding supply, even older chips get recycled. In 2024, cryptocurrency mining companies like Core Scientific and IREN pivoted to AI datacenters, activating a used GPU market. This suggests older hardware retains economic value.

Here's how I see it: These companies' accounting is "aggressive" within GAAP boundaries. Not fraud, but not conservative either. Burry's "scam" label is extreme, but his calculations are over 90% valid. Neutrally speaking, it's "a strategy to maximize profits within accounting standards."

Why These Companies? Burry's Strategic Targeting

What's interesting about Burry's criticism is who he attacked. He didn't mention pure AI companies like OpenAI, Anthropic (makers of Claude), or Cohere. Why?

First, pure AI companies are mostly private with no public financial disclosures. Burry's depreciation analysis relies on 10-K data, which doesn't apply to them.

Second, pure AI companies' profitability is already questioned. OpenAI's annual recurring revenue (ARR) is $3.4 billion, but it's posting massive losses due to model training costs and infrastructure maintenance. Anthropic is similar. Criticizing them would just elicit "of course they're losing money" responses.

The hyperscalers are different. They claim to be monetizing AI by integrating it into existing cloud, advertising, and software businesses. Looking at Q3 2025 results:

Company Quarterly Revenue (YoY Growth) Operating Income AI Contribution
Amazon $180B (+12%) $17.4B AWS +20%, AI recommendations
Microsoft $77.7B (+13%) Strong Azure +21%, Copilot productivity
Google $102B (+16%) $31.2B Cloud +34%, Gemini search
Meta $51.2B (+26%) $20.5B Ads +26%, Llama content
Oracle $14.9B (+12%) $6.2B Cloud +25%, Autonomous DB

These companies are showing real revenue growth from AI. But Burry asks: "Is this growth real, or an illusion created by depreciation tricks?" That's the crux. Burry's targeting is intentional. These five companies control 67% of the AI infrastructure market, so attacking them plants doubt about the entire AI sector.

And Burry currently holds put options on Nvidia and Palantir. If hyperscalers stumble, Nvidia's largest customer base takes a hit. Stock prices fall, and Burry's puts appreciate. Coincidence? I think not.

The November 25th Teaser: Fear Marketing 101

Burry's November 10th post ended with an ominous line: "More details on November 25th." This single sentence had significant market impact. Oracle dropped 5% immediately after, Meta fell 3%, and Nvidia's volatility spiked.

I see this teaser as "fear marketing." Burry didn't reveal specifics—just a vague promise of "more details." Why is this effective? Human psychology reacts more strongly to unknown threats.

Investors imagine: "What else does Burry know? Whistleblower evidence? A bigger accounting scandal?" This "anticipatory anxiety" is more powerful than actual announcements. And this anxiety increases volatility. Put option values rise with higher volatility of underlying assets. Burry knows this.

What will November 25th bring? I predict three scenarios:

  1. Concrete Evidence: Specific internal documents or additional calculations could emerge. This would be highly impactful.
  2. Expanded Claims: He might elaborate on existing points or add more companies. Impact would be limited.
  3. Disappointing Reveal: Weaker-than-expected content could trigger "was it all hype?" reactions. Stocks would rebound.

But regardless of which scenario plays out, Burry has already achieved his goal. He's maintained market anxiety for two weeks from November 10th to 25th, creating conditions favorable to his positions. That's a strategist's playbook.

My View: There's a Bubble, But It's Still Early

Having heard Burry's case, let me share my thoughts. I believe an AI bubble exists. But it's not the imminent collapse Burry seems to expect. Rather, it's still in early stages.

Why? First, AI is creating real value. Google's Gemini improves search quality, boosting ad click-through rates. Microsoft's Copilot increases enterprise customers' coding speed by 30-40%. Meta's Llama-based content recommendations extend user engagement time. These aren't illusions—they're measurable effects.

Second, AI infrastructure investment requires a long-term view. Consider the dot-com bubble. After the 2000 collapse, many companies disappeared, but Amazon and Google survived to become giants. Fiber optic cable and server investments were criticized as "excessive" then, but now we have the cloud era thanks to that infrastructure. AI is similar. Today's GPU investments might seem excessive, but in 5-10 years, new industries will be built on top of them.

Third, bubbles follow "boom and bust" cycles. AI is currently in the early boom phase. Using Gartner's Hype Cycle, we're past the "Peak of Inflated Expectations." The "Trough of Disillusionment" is coming. Some companies will fail, some investments will lose money. But afterward, the real winners will emerge on the "Slope of Enlightenment."

Burry seems to think the "Trough of Disillusionment" is imminent. I believe it's 1-2 years away. Why? Because companies are actually succeeding at AI monetization. As long as revenue growth continues, investment will too. And given AI's versatility, I expect these assets to remain productive even after they've moved past the cutting edge.

What Should Investors Believe?

So who should we trust? Burry? The companies? The market?

My answer is simple: Trust your own analysis. Burry's warnings and companies' earnings reports are all just "information." How you interpret that information and what conclusions you draw are your responsibility.

Burry's depreciation analysis is valid. But calling it "fraud" is hyperbole. Companies' accounting is aggressive but legal. And there's evidence that AI is generating real profits. Both are true. How do you weigh these conflicting facts?

Here's my approach: I search for necessary data and analyze it myself first. Then I analyze it again with AI, cross-checking the information and conclusions AI and I each have, and together we build new conclusions.

At the same time, I listen to short sellers like Burry. They're skilled at finding risks the market misses. But I also consider their motives. When Burry issues warnings while holding put options, it's not pure prophecy—it's strategic messaging.

Ultimately, investment decisions are mine to make. If I determine the AI bubble will burst in 2026, I reduce my positions. If I believe the bubble is still early-stage, I continue investing. But in any case, I take responsibility for my analysis. I don't blame Burry or resent companies.

Closing: Waiting for November 25th

On November 25th, Burry's promised "more details" will be revealed. I'm interested but also skeptical. He might drop a bomb, or it might be a disappointing repeat. Either way, I won't adjust my portfolio based on his words.

Instead, I'm treating this as a learning opportunity. I'll study Burry's analytical methodology, understand companies' counterarguments, and observe market reactions. And I'll refine my investment philosophy.

The AI bubble will eventually correct. Whether that's 2026 or 2028, I don't know. But even after that, the AI industry will continue growing. The real winners are those who endure short-term volatility and capture long-term value.

Burry plays his game. I play mine. I hope you'll play yours. Nobody is responsible for your money. Only you can be.

References:

  1. Michael Burry's X Post (11/10/2025): https://x.com/michaeljburry/status/1987918650104283372
  2. CNBC - Michael Burry accuses AI hyperscalers of boosting earnings (11/11/2025): https://www.cnbc.com/2025/11/11/big-short-investor-michael-burry-accuses-ai-hyperscalers-of-artificially-boosting-earnings.html
  3. Seeking Alpha - Michael Burry accuses tech giants of inflating profits: https://seekingalpha.com/news/4519876-michael-burry-accuses-tech-giants-of-inflating-profits-by-stretching-asset-lifespans
  4. Business Insider - Michael Burry's AI Warning and Jim Chanos: https://www.businessinsider.com/michael-burry-ai-stocks-nvidia-meta-oracle-hyperscalers-jim-chanos-2025-11
  5. Benzinga - Michael Burry Doubles Down On AI Bubble Claims: https://www.benzinga.com/markets/equities/25/11/48774428/michael-burry-doubles-down-on-ai-bubble-claims
  6. Sherwood News - Michael Burry's AI accounting concerns: https://sherwood.news/markets/michael-burry-has-some-concerns-about-ai-accounting/
  7. TipRanks - Michael Burry's $176B Depreciation Gap Warning: https://www.tipranks.com/news/michael-burry-sounds-alarm-on-176b-depreciation-gap-among-tech-giants

Comments

Post a Comment