Burry Added to His Nebius Short at $247. His Argument Is About Depreciation

Michael Burry added to his short position in Nebius at $247, and the argument underneath it is narrower and more specific than the headline quote suggests. He called the company "what the top of a boom looks like," which travels well and explains nothing. The reasoning is about accounting and pricing.

Start with the accounting. Nebius extended the depreciation schedule on its servers from four years to five. Stretching that schedule spreads the same hardware cost over more quarters, which lowers the depreciation charge in each one and raises reported earnings without a dollar of new revenue. Companies do this legitimately when equipment genuinely lasts longer. Burry's objection is that the business itself does not appear to believe the hardware lasts longer, and he points at the pricing to say so.

That is the second half. Customers pay $40 to $50 million per megawatt for short-term access to compute, against $20 to $25 million for medium-term commitments. Roughly double, for the privilege of not committing.

Read that spread as a market price on obsolescence. If buyers expected today's chips to stay competitive for years, locking in a long contract would be the cheap and obvious move, and the short-term premium would be thin. A premium that large says a meaningful set of buyers would rather pay twice as much than be married to this generation of silicon. Burry's read is that the business is counting on roughly 50% annual decay in the value of its deals, which sits badly beside an accounting change that assumes the opposite.

He is short more than Nebius: Micron, Oracle, and the SOXX semiconductor ETF are in the same book. So this is a position on the shape of the AI infrastructure trade, not a single-company forensic call.

Now the other side, because it deserves more than a sentence. The bull case does not require the hardware to last forever. It requires utilization and pricing to hold long enough to earn the capex back, and that is a genuinely open question rather than a settled one. Compute demand has run ahead of supply for two years. If it keeps doing that, a five-year schedule is defensible and the short-term premium is just what scarcity looks like, not a forecast of decay. Depreciation assumptions are estimates about the future, and Burry's estimate is not automatically better than the company's because it is more pessimistic.

The market has been emphatic about which side it takes. Nebius is up roughly 210%, and it rose 34% on an earnings beat. That is the tension in this story and it should not be smoothed over: the position is underwater, publicly, by a lot. Being early and being wrong produce the same brokerage statement for a long time, and only one of them eventually diverges.

What makes the call worth reading is that it is falsifiable. Watch whether that short-term premium compresses. Watch whether the depreciation change gets revisited. Those are observable, and they will settle this argument long before the share price does.