Hugging Face, the online home of the open AI model movement, is being absorbed by the chipmaker whose hardware powers much of it. Nvidia said September 3 that it will acquire the platform for $12.9B, and it promised that nothing about how developers use the service will be forced onto its own technology.
The deal keeps Hugging Face open, multi-cloud and multi-accelerator, with its brand intact. Founder Clement Delangue brought the idea to Jensen Huang himself, the announcement said, and the Nvidia chief framed the purchase as a bet that open-weight models will keep driving AI adoption.
Hugging Face’s numbers explain the price tag. More than 18M developers share over 3M models, 500K datasets and 1M applications through it, and more than 200K companies build on the platform. Nvidia already ranks as its most prolific corporate contributor, with more than 500 models and 250 datasets published there.
The timing is no accident. OpenAI and Anthropic are pressing Washington to regulate open-source releases, and Nvidia has joined Microsoft, Google and Meta in defending open weights in public letters. Owning the distribution point for those models lets Nvidia shape the evaluation and inference tooling around them while keeping its own hardware optional, a position that protects both its ecosystem and its antitrust posture.
