He Closed His Fund So He Could Say This

In November 2025, Michael Burry did something unusual for a man who had just been proven right about almost everything he worried about. He shut it all down. He deregistered Scion Asset Management — the fund that famously profited from the 2008 housing collapse and again through the dot-com bust — and stepped away from outside money. In his final letter to investors, he conceded that his read on value had drifted out of step with the market for some time. Translation: he had been early, and he knew it.

Days later he resurfaced on Substack under a name that tells you everything about how he sees his role: Cassandra Unchained — after the Greek prophet cursed to speak the truth and never be believed. Freed from the compliance rules he felt had muzzled him, he began publishing the most detailed bearish case of his career. The first post added tens of thousands of paying subscribers in a single day.

This matters for one reason. Burry is no longer talking his book to clients. He has no fund to defend. What he is publishing now is, by his own framing, simply what he believes — and what he believes is that the AI-driven market has stopped doing the one job a market exists to do.

Price Has Stopped Listening to News

Burry's plainest statement came after a long spring drive with nothing but the radio for company — hour after hour of one subject. A.I. Nothing else. He came home and wrote that stocks were no longer rising or falling on jobs or consumer sentiment.

The evidence was hard to wave away. A reading of consumer sentiment near a record low landed on the very same day the S&P 500 closed at a record high. When price stops responding to information and simply follows its own momentum, it has stopped discovering value. Burry said it felt like the last months of 1999.

"Not Enron. Clearly Cisco."

When Nvidia took the rare step of sending sell-side analysts a memo rebutting his work, Burry's reply drew a careful line:

The distinction is the whole argument. He is not alleging fraud at Nvidia, nor that AI fails to work. He is saying Nvidia is the picks-and-shovels supplier to a build-out that everyone is treating as permanent. Cisco built the plumbing of the late-1990s internet. Its revenue was real, its customers were real, the technology genuinely changed the world — and the stock still fell more than 80% from its peak when the buying slowed, and never regained those highs. Burry's claim is that the AI infrastructure boom rests on the same fragile assumption, and he lays out the mechanism in three parts.

The Argument That Made Nvidia Respond

Burry's most concrete charge isn't about Nvidia at all — it's about its customers' books. He alleges the hyperscalers are quietly inflating their own profits through depreciation. The logic is simple. When a company buys an expensive asset, it spreads the cost over the years it expects to use it. Stretch that estimated useful life, and you report a smaller expense each year — and a bigger profit.

Here is the tension. Nvidia ships a meaningfully faster generation of chips roughly every twelve months. Yet the companies buying them have been steadily extending how long they assume those chips will stay useful.

Burry's verdict was blunt: understating depreciation by stretching an asset's useful life is "one of the more common frauds of the modern era." He estimated the maneuver could understate depreciation across the sector by roughly $176 billion between 2026 and 2028 — and projected that by 2028 it could overstate Meta's profits by about 21% and Oracle's by about 27%. His point is about timing: extending useful lives front-loads profit today and back-loads the cost. If demand cools in 2026 or 2027, the writedowns could land all at once, erasing much of the AI earnings investors are now celebrating.

When a Blowout Quarter Isn't Enough

On May 20, 2026, Nvidia did almost everything a bull could ask. It reported record quarterly revenue of about $81.6 billion, up 85% from a year earlier, and beat Wall Street's estimate by nearly $3 billion. It guided next quarter higher. It authorized an $80 billion buyback and raised its dividend twenty-five-fold.

And the stock fell — roughly 1.5% that night, and another 4.3% the following week. When a company clears every bar that high and the shares still slip, it is worth asking what is already priced in. A market that cannot rise on great news is a market running on expectation rather than surprise — which is precisely the condition Burry has been describing.

A serious investor holds both possibilities at once: Burry's mechanism may be sound and the timing may stay wrong for a long while. That ambiguity is not a reason to ignore him. It is the reason to prepare rather than predict.