On September 3, 2026, NVIDIA announced it has agreed to acquire Hugging Face for $12,930,300,000. CEO Jensen Huang framed the purchase as scaling the open-model hub’s infrastructure while keeping the platform open across frameworks, clouds, and accelerators — and said NVIDIA compute will not be required to build or deploy through Hugging Face.
Earlier Brocker coverage tracked the talks while they were still unsigned: Nvidia closes in on Hugging Face acquisition for $12.9 billion.
Confirmed
- Agreed price: NVIDIA’s blog states an agreement to acquire Hugging Face for $12,930,300,000.
- Open-platform pledge: Hugging Face will remain open for model choice, frameworks, clouds, inference providers, and computing platforms; multi-cloud and multi-accelerator support continues.
- Platform scale (NVIDIA’s figures): more than 18 million developers/researchers/creators; more than 3 million models, 500,000 datasets, and 1 million applications; more than 200,000 companies on the platform.
- NVIDIA as contributor: NVIDIA says it is the largest contributor of open models and data on Hugging Face — more than 500 models and more than 250 open datasets released there.
- Open-weights advocacy: Huang notes a recent open letter he co-authored on the importance of open weights for access, cybersecurity, sovereignty, and distributed AI leadership.
- Leadership continuity signal: Huang says Clem Delangue approached him about Hugging Face’s next chapter and that the team will keep the brand while joining NVIDIA’s larger canvas.
Unknown
- Close and regulators: The announcement is an agreement to acquire — not a closed transaction. Timing, conditions, and any antitrust review are not spelled out in the blog post.
- Neutrality under pressure: The multi-accelerator pledge is clear today. How placement, evaluation defaults, and inference products behave when NVIDIA silicon competes with rivals is still an open operational question.
- Secondary deal math: Earlier reporting cited a ~$150M revenue run-rate, a prior $4.5B funding valuation, and a declined ~$500M NVIDIA investment at a $7B valuation. Those figures are not restated in NVIDIA’s agreement post — treat them as prior reporting context, not the new disclosure.
- Cloud-capacity thesis: Using Hugging Face endpoints to absorb unused committed cloud capacity remains an analyst angle from pre-agreement coverage, not a claim in the NVIDIA announcement.
Why it matters
Owning the central open-weight hub gives NVIDIA a distribution layer just as closed labs keep building custom silicon to reduce GPU dependence. A healthy open ecosystem is still mostly served on NVIDIA hardware; controlling the hub is a way to keep that flywheel spinning without forcing NVIDIA compute as a platform requirement — at least on paper. It also extends NVIDIA’s surface area from chips and libraries into the place developers already discover, evaluate, and deploy models.
Our take
The strategic logic is distribution, not charity: NVIDIA is buying the town square for open models after the price had already been floated in unsigned talks. Huang’s neutrality language is the right public contract for the community that made Hugging Face valuable. The live test is whether that contract still holds when NVIDIA-optimized stacks and rival accelerators fight for the same default slots. Until close, read this as a signed intent with a very large check — not a finished change of control.
