Michael Burry Is Short Nvidia’s Multiple, and His Options Expire First

On Nov 12 Michael Burry doubled a position in put options that expire in early 2027. The bet does not sit on one company. It runs through a semiconductor exchange-traded fund and a list of names he has been willing to say out loud: CoreWeave, Nebius, TSMC.

The comparison he keeps making is with 2000. He has described the Nasdaq 100 at its high as the same shape as the internet peak, and the point of that comparison is a multiple: what an investor pays for a dollar of earnings, against what those earnings have historically grown into.

That is the whole argument in one line. The AI buildout is real. The price of participating in it assumes the buildout continues at the current rate for longer than the contracts on the books support. A multiple is a claim about the future, and a short is a claim that the future is priced too high.

The semiconductor fund short is the broad version of the trade. The single names are the concentrated version. TSMC sits at the bottom of the chain, CoreWeave and Nebius at the top of the rented-compute layer, which means the position covers the companies that would feel a demand pause first and the ones that would feel it hardest.

His stated trigger is Jensen Huang’s departure. Burry’s argument is that the exit weakens both the margin structure and the certainty of growth at Nvidia, because both were built around a founder who set supply commitments and pricing personally. Take the founder out and the two things investors paid a premium for become assumptions rather than facts.

The timing is where the trade actually lives. Puts expiring in early 2027 give him three to five months from the day he added to them. His public call is a systemic correction in chip stocks before the middle of 2027. Those two dates do not have to line up, but if the correction arrives in the second half of 2027, the options are gone and the thesis is right.

Being early is the standard failure mode for this kind of bet. Carry costs money every month, implied volatility on chip names is expensive to own, and a put that expires worthless has to be paid for out of the next one. A correct call delivered through the wrong instrument is a loss.

The other side has a simpler argument. The companies he is shorting are still signing contracts, the largest buyers of computing have not cut their capital plans, and a correction that has been predicted more than once has not arrived. Every quarter that passes without one adds to the cost of holding the position.

What a multiple-based short needs is earnings estimates that break rather than stall. A stock at 20 times earnings that grows 20% is priced roughly right on that arithmetic. The same stock at 20 times that grows 5% is a de-rating candidate, and the de-rating arrives before the earnings do. The bet works only if the second case reaches the AI infrastructure chain.

He could be wrong on the specifics. Nvidia’s margin depends on product mix and supply costs as much as on who runs the company, and TSMC has carried pricing power through demand shocks worse than a pause in AI orders. A founder’s exit is a governance event that markets have historically priced in weeks, not years.

He could be right on the mechanics. The rented-compute names on his list carry the most operating gearing to a pause, because their revenue is contracted against capacity they have already committed to buy. A buyer who cancels does not cancel the lease underneath it.

For investors who are not running the trade, the useful part is the checklist rather than the position. Watch order intake at the bottom of the chain, utilization at the rented-compute layer, and capital spending guidance from the handful of companies large enough to move the whole market. Those three series turn before a stock does.

There is also a valuation comparison worth keeping separate from the narrative. A bubble call is a claim about the price relative to earnings, and it can be tested quarter by quarter as estimates move. The 2000 analogy is a mood, the multiple is the measurement, and only one of them can be checked.

The verifiable hypothesis is clean. If chip stocks correct before mid-2027, his window is right and the puts pay. If the correction arrives after the options expire, the call was directionally correct and financially wrong. The date to mark is early 2027, not the mid-year target.

Burry has doubled down on a date as much as on a direction. The market will settle the direction long before his options settle the date.

For anyone who agrees with the thesis but not the instrument, the practical version is smaller and slower. Trim exposure to the names with contracted capacity, keep the ones with pricing power, and let the next two earnings seasons do the work the puts were hired to do.

Related Posts

  • November 20, 2026
  • 17 views
Sandisk Completes Eight-for-One Split as NAND Outlook Darkens

Eight for one. Sandisk’s stock split took effect on Nov. 19, and the shares began trading on the adjusted basis on Nov. 20, a change that divides the price by…

  • November 20, 2026
  • 15 views
A Lawmaker’s Privacy Suit Puts Model Design on Trial

For three years British courts have sidestepped the question of whether the company that builds an AI model answers for what its users produce with it. A written defence filed…