NVIDIA paid $12.93bn for Hugging Face, not the ten billion being repeated

NVIDIA did not pay ten billion dollars for Hugging Face. It paid $12,930,300,000, and the figure has been sitting in the first sentence of Jensen Huang's own announcement since 3 September 2026.

On 9 September Nathan Lambert — who worked at Hugging Face, and whose read on the deal is the most informed one published so far — wrote this:

For Nvidia, who is so rich and powerful, this ability is worth easily the ~$10B they paid, but every year.

The argument survives the correction. The number does not. The real price is 29 percent above ten billion, and if you are sizing what NVIDIA thinks open-model distribution is worth, that is not a rounding error.

Why the wrong number stuck

Probably because ten is what a deal of this shape sounds like, and because nobody re-reads the first line of a blog post they have already summarised. We would gently suggest re-reading it. The price is exact to the hundred dollars, which is what a share-exchange calculation produces and what a negotiated headline number never does. Somebody divided a fixed pool of NVIDIA stock by a strike price and printed the result. The byline on the post, as the page renders it, is 黄仁勋.

So what does that buy?

Divide the price by the developer count and NVIDIA paid roughly $718 a head. That is our arithmetic, not theirs. For comparison it sits in the range enterprise software buyers pay for a seat that generates revenue, and Hugging Face developers mostly do not.

Which is Lambert's point, and it is the right one. The platform's asset is influence over where the open-model conversation goes, and influence has no line on a profit-and-loss statement. He argues NVIDIA is a better owner than any of the three big clouds precisely because it feels no pressure to make the thing profitable, and can instead point it at winning "the hearts and minds of the next 100 million AI developers".

Does the difference actually matter? If you accept Lambert's valuation — that the platform is worth the whole purchase price to NVIDIA every year — then no, and three billion dollars is a detail on a balance sheet that carries far larger ones. That is a fair objection and we would not argue with it hard.

But prices get quoted for years after the deal closes, in comparisons, in board decks, in the next negotiation over what an open-model registry is worth. Getting it wrong now means getting every ratio built on it wrong later. The round number also hides what the exact one shows, which is that this price was computed rather than negotiated to a headline, and computed prices tend to move with whatever was on the other side of the division.

The neutrality clause is the whole asset and the whole risk

Huang commits in the post that developers will choose their own models, frameworks, clouds and inference providers, and that "NVIDIA compute will not be required to build on or deploy through Hugging Face". NVIDIA is already the platform's largest contributor of open models and data, which is either reassurance or the problem, depending on how the next year goes.

A registry everyone trusts is worth twelve billion dollars. A registry that quietly routes to one vendor's silicon is worth a fraction of that, because the moment it tilts, the second-best registry becomes free.

So watch the deploy button on a model page rather than the press release. If a year from now the default inference path on huggingface.co has a green tint, the thesis Lambert wrote and the promise Huang made will have come apart — and the number worth remembering is the one NVIDIA printed itself.