Nvidia acquires Hugging Face for $12.93 billion
Nvidia confirmed on 3 September 2026 that it had agreed to acquire Hugging Face for $12.93 billion. The platform hosts roughly three million models, half a million datasets and one million applications, and is used by more than 18 million developers and over 200,000 companies. Jensen Huang said Hugging Face would remain an open platform for the entire ecosystem, that developers would keep their choice of models, frameworks, clouds and inference providers, and that Nvidia compute would not be required to build on or deploy through it. It is Nvidia's second-largest acquisition, after the roughly $20 billion purchase of Groq assets at the end of 2025.
Why It Mattered
The open side of machine learning has had a single de facto commons: the place where weights, datasets, model cards and evaluation artefacts are published and retrieved. That commons is now owned by the company supplying most of the hardware the field runs on. The commercial logic is ordinary — integration up the stack, a hedge against any slowing of chip demand — but the structural consequence is not. Distribution, rather than training, is where open-weight ecosystems are governed in practice: what may be hosted, what is gated, whose licences are enforced, which formats and runtimes are first-class, which hardware backends get day-one support. Nvidia's neutrality commitment is unusually specific, and it will be measured against the incentives of an owner that earns more when models are large and inference is expensive. The entry is also a marker of where value settled in this period: an accelerator vendor paying thirteen billion dollars for a repository rather than for a model or a laboratory, on the same day a frontier lab shipped a system at critical cyber capability. Competition review is the open question. An acquisition that places the main open-model channel under the dominant accelerator supplier invites scrutiny in both Washington and Brussels, and any conditions imposed — on hosting neutrality, on hardware-agnostic support, on data access — would shape how open-weight AI is distributed well beyond this decade. If the platform stays neutral, this will read as consolidation without capture; if it does not, it will read as the moment open AI acquired a landlord.
Who Built It
Nvidia
Applications
- Open Model Distribution
- Developer Infrastructure