Nvidia is buying Hugging Face to bind millions of AI developers more closely to its chips, as hyperscalers including Amazon and Google race to build custom processors that could bypass it.
The world's most valuable chipmaker confirmed on Thursday that it has agreed to acquire the open-source AI platform for approximately $12.9 billion.
Defending the ecosystem
Hugging Face operates a platform often compared to GitHub, where developers share and download AI models, datasets and applications rather than traditional software code.
Analysts have likened the move to Microsoft's 2018 purchase of GitHub, which used ownership of a widely used developer hub to steer programmers towards its own cloud services.
By controlling Hugging Face, Nvidia gains direct access to more than 18 million developers, strengthening the pull of its hardware and software even as competitors develop chips designed to reduce reliance on it.
The platform hosts over three million models, 500,000 datasets and one million applications, according to Nvidia.
A steep multiple
Hugging Face's revenue is hard to pin down precisely, but estimates from Business Insider and The Information put annualised revenue at somewhere between $100 million and $150 million.
At $12.9 billion, that implies a multiple of roughly 86 to 129 times sales, an extraordinary premium even by the standards of the current AI boom.
The price also values Hugging Face at nearly three times the $4.5 billion it commanded in its 2023 funding round, a deal in which Nvidia itself was already an investor.
Chief executive Clément Delangue has said the company is close to profitability, having only recently begun spending the capital raised three years ago.
Will it pay for itself
On a standalone basis, a multiple in the high double digits or low triple digits of revenue would take Hugging Face many years, if ever, to grow into on subscription and enterprise fees alone.
The strategic case, however, rests on a different calculation: Nvidia is not primarily buying a software business but the central distribution point through which millions of developers discover and deploy AI models.
Owning that funnel makes it easier for Nvidia to keep its chips as the default choice for running those models, and to sell spare data centre capacity to enterprises through Hugging Face's platform.
Huang has framed the deal partly in defensive terms, saying open-source AI gives defenders against cyberattacks an "asymmetric advantage" over attackers, a point he tied to a recent security breach at Hugging Face.
Whether that strategic value ever shows up as a direct return depends less on Hugging Face's own revenue and more on whether the deal measurably slows the drift of developers towards custom chips built by Amazon, Google and Microsoft.
Deal terms
According to a regulatory filing, the transaction comprises an $11.9 billion purchase price for shareholders plus up to $1 billion in equity-based retention payments for Hugging Face staff who join Nvidia.
Delangue and co-founders Julien Chaumond and Thomas Wolf will move across to Nvidia as part of the agreement, and the deal is expected to close in the first half of 2027, subject to regulatory approval.