Michael Burry and the AI bubble: what he argues
The essentials in 30 seconds
- Michael Burry argues that AI hardware accounting makes hyperscaler earnings look stronger than they are.
- His fund’s filing listed puts with a notional value of $187 million on Nvidia and $912 million on Palantir (CNBC, November 2025).
- He estimated about $176 billion of understated depreciation from 2026 through 2028.
- CNBC could not confirm the practice independently, so his figures stay estimates.
- Three deeper pages cover the Big Short comparison, the puts, and depreciation.
Michael Burry is the investor most often quoted on the AI bubble, and his case is more specific than a warning. He points at positions, accounting choices, and a pattern he knows from 2008. As of October 2026 his claims are partly documented and partly estimates. This page sits under the guide on who says AI is a bubble, sorts the documented from the estimated, and routes you to three detailed pages.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
Why do investors listen to Michael Burry on the AI bubble?
Burry runs the fund Scion Asset Management and became widely known for betting against US mortgage bonds before the 2008 crisis. That record explains the attention. It also sets a trap: one correct call does not make the next one correct, and a bubble call needs its own evidence. This page treats his record as a reason to read him, never as proof.
His AI bubble case has 3 parts that you can check separately: positions he filed, an accounting argument he posted, and a historical comparison he invites. Each part has a different level of proof, so a reader can accept one and question another without rejecting the whole view. The rest of this page follows that order, from the best documented item to the least documented one.
What is Michael Burry’s core argument about the AI bubble?
His core argument is that AI spending produces reported profits that look better than the cash reality. Chips bought today lose value fast, yet companies spread their cost over many years. If the spread is too long, yearly costs look small and earnings look large, which supports high share prices. Prices then rest on a number that a change of assumption could reduce.
This is an argument about measurement, and measurement can be tested against filings. That makes it more useful than a prediction about a crash date. You can follow the argument without taking a view on timing, and you can recheck it each quarter when companies report new figures.
How does Michael Burry’s Big Short compare with the AI bubble?
The comparison with 2008 is the part of his case that draws the most debate, because the two markets differ in structure. The mortgage trade targeted a debt instrument with a visible weak point. The AI case targets equities and accounting estimates. Some features repeat, such as a widely held asset and a belief that it cannot fall.
The dedicated page, Michael Burry’s Big Short and the AI bubble, tests the comparison point by point: the instrument, the weak link, the cost of waiting, and the exit. Reading it first helps you judge how much the 2008 record should weigh in the AI case.
What do Michael Burry’s puts on AI stocks actually show?
A filing for 30 September 2025 listed put options with a notional value of about $187 million on Nvidia and $912 million on Palantir (CNBC, November 2025). Notional value measures the stock the options cover. It does not measure the money paid or the profit at stake, and the filing gave neither strike prices nor expiry dates.
So the filing shows a direction and a size at one date. It does not show whether the positions were a hedge, were closed, or changed after that date. The page on Michael Burry’s puts on AI stocks explains how to read option filings and what to conclude from this one.
Why does Michael Burry say AI depreciation hides the AI bubble?
On 10 November 2025 Burry wrote on X that extending the useful life of chips and servers understates depreciation and inflates earnings. He said Nvidia products follow a 2 to 3 year cycle. He estimated about $176 billion of understatement from 2026 through 2028, and profit overstatement of roughly 27% at Oracle and 21% at Meta by 2028 (CNBC).
CNBC could not independently confirm the practice, and Nvidia declined to comment. Accounting rules give companies room to set useful lives, which makes the claim hard to prove or to refute. Michael Burry’s depreciation argument about the AI bubble has its own page, with two deeper pages on chip lifespans and company schedules.
Where does the Michael Burry case on the AI bubble fall short?
Three limits apply. The figures are his estimates, not company disclosures. The put filing is a snapshot with missing terms. And a correct view on accounting can still lose money on timing, because prices can stay high for years. A sound reading keeps these limits next to the claims, so a good argument does not turn into a trading signal.
His stake matters as well. A fund that holds puts gains when prices fall, which is a reason to check his numbers against filings. This applies to every voice in the series, as the guide on who says AI is a bubble explains, and it never makes a claim false.
How can you use the Michael Burry argument to test the AI bubble?
Turn the argument into 3 checks. Read the useful-life note in each annual report of the AI spenders you follow. Recompute earnings with a shorter life. Compare the gap with total profit. A large gap supports his view, and a small one weakens it. The method works with public filings and a spreadsheet, and it takes an afternoon per company. Write down the result with its date, so next quarter you can see whether the gap widens or closes.
This page is general information as of October 2026, not investment advice. If AI spending sits inside a business decision of yours, the EQOS Asia pricing page lists the diagnostic formats that score your exposure with a reproducible method instead of an opinion.
See also in this guide
Frequently asked questions
What is Michael Burry’s argument about the AI bubble?
He argues that AI hardware loses value faster than hyperscalers’ accounting shows. Long useful lives make yearly depreciation small and earnings large. He estimated about $176 billion of understated depreciation from 2026 through 2028. CNBC could not confirm the practice, so the figure stays an estimate.
What positions did Michael Burry file on AI stocks?
A filing for 30 September 2025 listed put options with a notional value of about $187 million on Nvidia and $912 million on Palantir. The filing gave no strike prices or expiry dates, so it is unclear whether the positions remain open or acted as a hedge.
Is the 2008 Big Short a good guide to the AI bubble?
It is a useful comparison for method and a weak one for timing. The 2008 trade targeted a debt instrument with a visible weak link. The AI case targets equities and accounting estimates, so it needs separate evidence and does not inherit the earlier result.
Does Michael Burry’s call prove an AI bubble?
A call is information about a risk. Only hindsight confirms a bubble. Burry’s stake as a put holder is a reason to verify his numbers against filings. This page is general information as of October 2026, not investment advice.
Test your knowledge
1. What does notional value measure in an options filing?
- The profit already earned
- The premium paid
- The stock covered by the options
Show the answer
The stock covered by the options.
2. What depreciation understatement did Burry estimate for 2026 through 2028?
- About $176 billion
- About $17 million
- About $1.76 trillion
Show the answer
About $176 billion.
3. Which firm did Burry say could see profit overstated by about 27% by 2028?
- Nvidia
- Oracle
- Palantir
Show the answer
Oracle.
4. Why can a correct accounting view still lose money?
- Filings are secret
- Puts never expire
- Prices can stay high for years
Show the answer
Prices can stay high for years.
5. Which check supports Burry’s view if it shows a large gap?
- Recomputing earnings with a shorter useful life
- Counting social media posts
- Comparing company logos
Show the answer
Recomputing earnings with a shorter useful life.
About the author
Stéphane Jambu
President, EQOS Forecast Asia
Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile
Sources: EQOS Asia pages (eqos-asia.com/ai-bubble/ and eqos-asia.com/pricing.html), CNBC report on Michael Burry as syndicated at https://www.albanyadvertiser.com.au/business/cnbc/big-short-investor-michael-burry-accuses-ai-hyperscalers-of-artificially-boosting-earnings-c-20655611, consulted on 22 October 2026. Page updated on 22 October 2026.