Nvidia said Thursday it has agreed to buy Hugging Face, the open-source AI platform used by millions of developers to share machine-learning models, for roughly $12.9 billion — one of the largest acquisitions in the chipmaker's history and its clearest bet yet that owning the software layer around AI matters as much as owning the silicon underneath it.
Under the terms described by the companies, Nvidia will pay about $11.9 billion to Hugging Face shareholders plus up to $1 billion in equity set aside to retain Hugging Face employees who join Nvidia, bringing the total value of the deal to nearly $13 billion. It is Nvidia's second-largest purchase after the roughly $20 billion deal it struck in December for assets of AI chip startup Groq, and it comes as Nvidia's market value has swelled on demand for the processors that train and run large language models.
An emoji-named startup becomes AI's default hub
Founded in New York in 2016 by French entrepreneurs Clément Delangue, Julien Chaumond and Thomas Wolf, Hugging Face began as a teenage chatbot app before pivoting to build what is now widely described as the "GitHub of AI" — a repository where researchers and companies post and download machine-learning models instead of building them from scratch. The platform now says it hosts more than 3 million models, 500,000 datasets and 1 million applications, used by roughly 18 million developers, researchers and creators and about 200,000 corporate customers. Nvidia was already among its investors, alongside Amazon, Intel, AMD and Salesforce, through a 2023 funding round that valued the startup at $4.5 billion.
Hugging Face chief executive Clément Delangue told CNBC that his company approached Nvidia chief executive Jensen Huang over the summer, concluding that Hugging Face and open-source AI needed more scale and resources to keep pace.
Hugging Face and open-source AI in general was at the turning point, and that it needed more, more resources, more scale, more visibility.
Clément Delangue, Hugging Face chief executive, to CNBC
In a blog post announcing the deal, Huang wrote that Hugging Face "will remain an open platform for the entire AI ecosystem," and Nvidia said its hardware would not be required to build on or deploy through the platform — an attempt to reassure a developer community that includes rivals such as AMD, Google and Amazon, all of whose chips are used to run models hosted on Hugging Face.
A deal that tests the limits of antitrust enforcement
The acquisition immediately drew scrutiny from analysts and commentators who argue that Nvidia, already the dominant supplier of AI training chips, would be taking ownership of the platform that has become the default distribution layer for open-source AI. Trade outlet The Register argued the platform is "too important to fall into Nvidia's hands," while Nvidia has described the deal as a "deconcentration platform" that widens access to AI tools rather than narrowing it. The deal is expected to close in the first half of next year, giving antitrust regulators in the U.S. and Europe months to weigh in.
Markets, at least initially, treated the deal as a positive for Nvidia. Shares of the company rose about 2.5% Thursday afternoon to roughly $227, up from a prior close near $224, as the S&P 500 and Nasdaq Composite each climbed more than 1% on the day. Investors have grown accustomed to Nvidia expanding well beyond chip design — the company has taken minority stakes in AI labs including OpenAI and Anthropic and made a string of smaller acquisitions over the past two years — but the Hugging Face deal marks one of its most direct moves into owning developer infrastructure outright.
For Hugging Face's roughly 200,000 corporate customers and millions of individual users, the immediate practical changes are likely to be modest: the platform is expected to keep operating largely as it has, hosting models built on rival chipmakers' hardware alongside Nvidia's own. The larger question, several analysts said, is whether that neutrality survives once Hugging Face's balance sheet, roadmap and incentives sit inside a company that profits from every AI workload steered toward its own processors — and whether regulators conclude the arrangement changes competitive dynamics across the AI supply chain enough to intervene before the deal closes.
