NVIDIA agrees to acquire Hugging Face and says the platform will remain open

NVIDIA announced on September 3 that it had agreed to acquire Hugging Face for $12.93 billion. The significance for open-weight AI is difficult to overstate: Hugging Face is not merely another model host, but a central distribution layer for model weights, datasets, demos and developer tooling. NVIDIA says the platform serves more than 18 million developers and hosts more than three million models, making any ownership change relevant to how the open-model ecosystem discovers and distributes artifacts.

The announcement is unusually explicit about openness. NVIDIA says Hugging Face will remain an open platform where developers can choose their models, frameworks, clouds, inference providers and computing platforms. It also states that NVIDIA compute will not be required and that multi-cloud and multi-accelerator development will continue. Those are commitments in NVIDIA’s announcement; they should be tracked over time rather than treated as evidence about future implementation before it happens.

For Open Model Weights, the practical rule is continuity of evidence. A Hugging Face repository should still be described as the source repository unless publisher ownership is separately established, and a platform-level acquisition should not silently change the developer identity, license or provenance of any model record. The news matters because infrastructure incentives can shape discovery, evaluation and deployment, but the registry’s job remains narrower: record what each source actually exposes, retain revisions and keep claims tied to evidence rather than corporate assumptions.