Why Michael Burry Shorted Parts of the AI Market
Michael Burry (Scion Asset Management) is famous for being early—and painfully contrarian. In late 2025, he drew headlines again by taking bearish options positions tied to two stocks many people associate with the AI boom: Nvidia and Palantir.
That naturally raised the question:
Is Burry betting on an “AI bubble” popping—or is he hedging in a frothy market?
The answer is: it can be both, and his public commentary points to a broader capex-cycle + valuation thesis, not just a simple “AI is fake” argument.
What Burry actually did (and why headlines can be misleading)
Burry’s fund disclosed put options on Nvidia and Palantir in regulatory filings (a put tends to gain value if the underlying stock falls). Reports described the positions as totaling around $1.1B in puts at the end of Q3 2025.
Important nuance:
A 13F is a snapshot of holdings at quarter-end, not real-time. Positions may change quickly.
Options positions can be tactical, short-term, or part of broader hedging strategies.
So this strongly suggests a bearish or protective stance at that moment, but it does not prove a permanent, all-in “AI collapse” bet.
The “AI bubble” case: what Burry seems to be warning about
1) The capex-cycle pattern: markets peak before spending peaks
Burry has repeatedly pointed to a familiar historical pattern:
Markets often peak in the middle of a boom, while capital spending keeps rising for another year or two—then drops hard.
That matters in AI because the industry is deep into an infrastructure buildout phase:
data centers
GPUs
power and cooling
networking
If markets decide returns aren’t arriving fast enough, stock prices can fall before spending slows.
2) Depreciation and overinvestment risk
Another recurring theme in Burry-related commentary is skepticism about how markets price companies spending massive sums on AI infrastructure.
AI hardware:
depreciates quickly
becomes obsolete fast
requires continuous reinvestment
If revenue growth lags expectations, the investment math deteriorates rapidly.
3) Valuations became crowded and consensus-driven
Burry’s reported bearish exposure focused on two of the most visible AI-narrative stocks:
Nvidia as the “picks and shovels” infrastructure winner
Palantir as a high-profile enterprise AI software story
When a trade becomes consensus, even excellent companies can be priced as if:
growth will remain perfect
margins will never compress
competition won’t matter
Burry has historically targeted crowded trades, not bad companies.
4) Circular incentives inside the AI ecosystem
A subtle but important concern is how parts of the AI ecosystem reinforce themselves:
cloud providers fund AI companies
AI companies buy cloud compute
chip vendors benefit from both
Even without fraud, these feedback loops can amplify booms—and reverse quickly if confidence breaks.
Why Nvidia and Palantir specifically?
A simple interpretation:
Nvidia represents peak optimism around AI infrastructure margins
Palantir represents peak optimism around enterprise AI adoption narratives
If you want to express skepticism about pricing, not about AI technology itself, these stocks become symbolic targets.
This does not mean “AI won’t matter.”
It means expectations may be ahead of reality.
The alternative interpretation: this could be hedging, not ideology
It’s important not to overread filings:
options can be short-dated
positions can change rapidly
puts may hedge other exposure
So the fair conclusion is:
Burry appears skeptical of AI-related valuations and the capex cycle—but the exact timing and magnitude of his bet are not public.
What “Burry being right” would actually look like
Not an AI apocalypse.
More likely:
AI stocks reprice downward
infrastructure spending slows after a lag
weaker AI startups shut down or consolidate
stronger players survive at lower multiples
This is the classic boom → overbuild → correction → consolidation cycle.
What this means for founders building AI products
The lesson is not “don’t use AI.”
It’s:
don’t build businesses with uncontrolled AI costs
don’t promise unlimited usage without real unit economics
treat AI as infrastructure, not magic
Companies that:
meter usage
route models intelligently
cache aggressively
understand cost per feature
are far more resilient in a downturn.
Bottom line
Michael Burry’s AI short looks best understood as a valuation, crowding, and capex-cycle bet—not a rejection of AI itself.
He appears positioned for a scenario where:
markets reprice AI leaders
spending stays high briefly, then slows
the AI narrative collides with financial reality
That kind of reset wouldn’t kill AI.