Michael Burry has never approached markets the way most hedge fund managers do. Before he became famous for calling the 2008 subprime meltdown, he trained as a doctor, not as an investment banker.
Key Takeaways
Michael Burry is betting against Nvidia, arguing that AI hardware is being depreciated far too slowly.
He believes big tech’s long depreciation schedules inflate reported profits during the AI boom.
Auditors and GAAP rules allow these practices, meaning the issue is structural rather than outright fraud.
After leaving medicine behind, he built Scion Capital and gained global recognition for spotting structural problems others ignored.
Now he has a new target: the foundations of the artificial intelligence boom — and specifically, the way companies account for the massive spending powering it.
Burry’s Big Concern Isn’t Nvidia’s Valuation — It’s the Math Behind AI Spending
A recent regulatory filing revealed that Burry purchased put options on Nvidia, signaling a bearish view on the world’s most important AI semiconductor company. But unlike typical bearish calls, his skepticism is less about Nvidia trading at a premium and more about how its biggest customers record the costs of the hardware they buy.
At the core of his thesis is depreciation — the accounting process that spreads the cost of servers, GPUs and networking gear over the number of years those assets are expected to remain useful.
In theory, depreciation schedules should match reality. In practice, Burry argues, the numbers do not line up.
The Product Cycle Has Shrunk — But the Accounting Hasn’t
Cloud giants like Microsoft, Google, Amazon, Oracle and Meta are pouring billions into Nvidia hardware to fuel their AI ambitions. However, they typically depreciate these purchases over three to five years.
Burry believes this timeline has become outdated.
Nvidia releases new GPU architectures at an extraordinarily fast pace — about every year and a half. Each new generation makes the previous one less attractive, and in many high-performance workloads, effectively obsolete.
If equipment only offers peak usefulness for two years, Burry argues, but companies expense it over twice that timeframe, the result is predictable:
profits appear higher than they actually are.
This discrepancy, he warns, creates the illusion of stronger margins and more efficient AI operations than reality supports.
Is Burry Calling Out Fraud? Not Exactly — But He Sees a Systemic Blind Spot
Burry has used strong language to describe what he believes is happening, suggesting a coordinated overstatement of technological usefulness across big tech. But most analysts say the issue is not fraud so much as a lag in adapting accounting rules to a rapidly evolving industry.
Public companies follow GAAP standards, and those rules allow management considerable discretion when determining useful life estimates. Depreciation is also a non-cash expense, meaning the timing differences affect reported earnings far more than a company’s actual ability to generate cash.
That nuance matters. Major tech companies also undergo rigorous audits from firms like Deloitte, EY, PwC and KPMG. Any blatant misrepresentation of asset lifespans would raise red flags long before reaching regulators.
What Investors Should Take Away From Burry’s Thesis
Although Burry’s warning is dramatic, it highlights a real tension: the AI boom is unfolding so quickly that financial reporting frameworks sometimes struggle to keep pace. If GPU cycles continue accelerating, depreciation practices may eventually need to change, potentially lowering reported earnings for the biggest AI investors.
That possibility doesn’t invalidate AI’s long-term potential — but it does challenge the glossy near-term narrative around profitability.
For now, Burry is once again placing a contrarian bet on an assumption he believes the market has overlooked. Whether this challenge to the AI accounting model becomes another “Big Short” moment is something investors will be watching closely.
