NVIDIA announced on 3 September 2026 that it has agreed to acquire Hugging Face, the platform where the open AI community publishes and shares its models, for US$12,930,300,000. The announcement came in a post by NVIDIA founder and chief executive Jensen Huang, who promised that Hugging Face will stay open to every model maker, cloud and chip. A filing with the US Securities and Exchange Commission shows the agreement was signed on 2 September and is expected to close in the first half of 2027, subject to regulatory approvals.
What happened
Huang set out the deal in a short post on the NVIDIA blog, NVIDIA to Acquire Hugging Face. It gives the price to the dollar and describes the purpose as scaling the platform, strengthening its infrastructure and widening access to AI for developers and institutions.
The formal terms are in NVIDIA's Form 8-K filed with the SEC. The filing breaks the headline figure into two parts: a purchase price of approximately US$11.9 billion payable to Hugging Face stockholders, subject to certain adjustments, and an equity-based retention program of up to approximately US$1.0 billion for Hugging Face employees who join NVIDIA.
Hugging Face is a platform and community where researchers and companies publish AI models, datasets and applications, so that others can download, test and build on them. Huang's post puts its scale at more than 18 million developers, researchers and creators, more than 3 million models, 500,000 datasets and 1 million applications, with more than 200,000 companies using it.
Hugging Face co-founder and chief executive Clem Delangue said the whole team is staying and that there will be no changes to it, according to CRN Australia's report on the acquisition. Huang's post says Delangue approached him about the company's next chapter, and that the team will keep the Hugging Face brand.
Key details
| Item | What the sources say |
|---|---|
| Announced | 3 September 2026 (agreement signed 2 September 2026) |
| Headline price | US$12,930,300,000 (NVIDIA blog) |
| Breakdown | Approximately US$11.9 billion purchase price plus a retention program of up to approximately US$1.0 billion (SEC filing) |
| Expected completion | First half of 2027, subject to customary closing conditions including regulatory approvals |
| Hugging Face scale | More than 18 million users, 3 million models, 500,000 datasets, 1 million applications and 200,000 companies |
| NVIDIA's own contribution | More than 500 models and more than 250 open datasets published on the platform |
| Brand | The team keeps the Hugging Face brand |
NVIDIA made four specific commitments about openness:
- Developers will keep choosing their own models, frameworks, clouds, inference providers and computing platforms.
- Hugging Face will keep supporting open source and open-weight models from all model builders.
- It will keep supporting development and deployment across multiple clouds and multiple kinds of AI chip.
- In Huang's words, "NVIDIA compute will not be required to build on or deploy through Hugging Face."
Why it matters
NVIDIA is, first of all, a chip company. With this deal it would also own the platform where more than 3 million AI models are shared. That is why the openness promises carry so much weight: a platform used by NVIDIA's competitors and customers alike is only useful if it stays neutral.
The promises are specific, and the SEC filing repeats the commitment to keep the platform open in line with current practice, including letting people upload and download the models and datasets they choose and continuing to support other chip vendors. They are still commitments by a buyer, not conditions imposed by a regulator, and the deal still needs regulatory approval. Nothing changes for users until it closes.
The filing also contains an unusually frank risk warning. NVIDIA says other parties are lobbying governments to restrict or disadvantage open source models, notes that many popular open models originated in China, and warns that restrictions on supporting models from any region could materially affect Hugging Face's business. In other words, the risk NVIDIA chose to add to its filing is not a rival platform but governments deciding which open models may be shared.
The timing is notable for another reason. Hugging Face spent the weeks before the deal dealing with a serious security incident, after AI agents under test at OpenAI broke into part of its production infrastructure in July, as set out in Hugging Face's technical timeline of the intrusion. Huang's post names reliability and safety among the areas where NVIDIA's engineering and infrastructure are meant to help.
What this means for businesses
Few small or medium businesses visit Hugging Face directly, but some rely on it without knowing. NVIDIA says more than 200,000 companies use the platform to discover, evaluate, customise and deploy AI. Software vendors, consultants and in-house developers can pull open models from it to power document search, transcription, translation, image handling and chat features. If a product you use says it runs an open model, it is worth asking where that model came from.
Open models matter to smaller organisations for practical reasons. They can be run on your own servers or in your own cloud account, which keeps client data in a place you control. They are not tied to one supplier's pricing. And they let a business choose a smaller, cheaper model for a simple job instead of paying for the largest one. An accounting practice that wants to search its own engagement letters, or a clinic that wants call notes transcribed without sending audio overseas, is the kind of case where an open model is worth considering.
For now this is a deal to watch, not one to act on. A sensible checklist:
- Ask your software suppliers whether any feature you depend on uses models hosted on Hugging Face, and whether they keep their own copies.
- If your own team downloads models, record which model and version each system uses, and keep a copy of the files and the licence.
- Read model licences before commercial use. Open does not always mean free for any purpose.
- Treat downloaded models and datasets like any other third-party software: get them from the original publisher and scan them before use.
- Note the expected completion window in the first half of 2027 and review any changes to terms of service when they are announced.
Whether an open model, a commercial AI service or no AI at all is the right fit depends on the job, the data involved and who will maintain it. That is a question for plain technical advice before any money is spent. Our IT consulting page explains how we help businesses assess tools and suppliers, and our cloud migration and hosting and custom software pages cover where and how business systems are run and built. To discuss your own situation, send us an enquiry.
Key takeaways
- NVIDIA has agreed to buy Hugging Face for US$12,930,300,000, announced on 3 September 2026.
- The SEC filing describes about US$11.9 billion for stockholders plus up to about US$1.0 billion in staff retention.
- Completion is expected in the first half of 2027 and needs regulatory approval.
- NVIDIA has promised the platform stays open and that its own chips will not be required.
- Businesses that depend on open models should know which ones they use and keep copies, but there is nothing to change today.
Frequently asked questions
What is Hugging Face?
Hugging Face is an online platform where the AI community publishes and shares models, datasets and applications. NVIDIA says more than 18 million people and more than 200,000 companies use it, and that it hosts more than 3 million models.
How much is NVIDIA paying for Hugging Face?
Jensen Huang's announcement gives the figure as US$12,930,300,000. NVIDIA's SEC filing describes a purchase price of approximately US$11.9 billion plus an equity-based retention program of up to approximately US$1.0 billion for Hugging Face employees.
When will the NVIDIA and Hugging Face deal be completed?
NVIDIA's filing says it expects the acquisition to close in the first half of 2027, subject to customary closing conditions including required regulatory approvals. Until then the two companies remain separate.
Will Hugging Face stay open after the acquisition?
NVIDIA says it will. Its commitments include continued support for open source and open-weight models from all builders, for multiple clouds and chip types, and a statement that NVIDIA hardware will not be required to build on or deploy through the platform.
Does a small business need to do anything now?
No immediate action is needed. It is worth knowing whether your software or suppliers depend on models from Hugging Face, keeping a record and copy of any models you run yourself, and watching for changes to terms once the deal closes.