Nvidia to Buy Hugging Face in a $12.93 Billion Deal

Nvidia Is Buying the Place Where Open Models Live

Nvidia entered a definitive agreement on September 2, 2026 to acquire Hugging Face, and announced it publicly on September 3. The price is $12.93 billion, making it Nvidia's second-largest acquisition ever, behind the roughly $20 billion it paid for Groq assets at the end of last year. The transaction is expected to close in the first half of 2027, subject to customary conditions including regulatory approvals.

• Price: $12.93 billion, with the deal reportedly including a $1 billion retention package for Hugging Face employees
• Scale acquired: more than 18 million developers and researchers, over 3 million models, and 200,000 enterprise customers
• Timeline: definitive agreement September 2, 2026; expected close in the first half of 2027
• Commitment: Nvidia says Hugging Face will remain an open platform for the entire AI ecosystem

How the Deal Came Together

Hugging Face CEO Clement Delangue approached Jensen Huang over the summer, according to both executives speaking to CNBC on the morning of the announcement. Delangue's framing was that open source AI had reached a turning point and needed more resources, scale and visibility than the company could supply alone. Huang's public line was that open models matter greatly to Nvidia.

That is a vendor's account of a vendor's motives, and it should be read as such. But the structural logic is not hard to follow from the outside. Hugging Face is where open weights get published, downloaded and benchmarked; Nvidia sells the hardware those weights run on. The company has been in the ecosystem for years — the two collaborated on the LeRobot open robotics stack back in July.

What Does This Mean for Open Source AI?

The question everyone asked within an hour of the announcement is whether the hub stays neutral. Nvidia's stated commitment is that the platform remains open and consistent with existing practices, and that the companies will together scale the platform, strengthen its infrastructure and expand access for developers and institutions.

Worth keeping in perspective: Hugging Face was never a neutral public utility in the legal sense. It is a venture-backed company founded in 2016 by Clement Delangue, Julien Chaumond and Thomas Wolf, and it has always had to fund itself. The practical question is not neutrality in the abstract but whether models from Nvidia's competitors continue to get first-class treatment on the hub, and that is answerable only by watching what ships over the next year.

The bull case is straightforward and real: infrastructure costs money, and a hub serving 3 million models with free bandwidth to 18 million users has a bill that scales badly. Nvidia can absorb that. The WebGPU kernel work that made browser inference 2.57x faster earlier this week is exactly the kind of engineering that gets easier with a much larger balance sheet behind it.

Reading the Price Tag

At $12.93 billion, this is not a bolt-on. It sits in the same tier as the largest AI infrastructure deals of the cycle, and it is a distribution purchase rather than a technology purchase — Nvidia is not short of engineers, it is buying the default place developers go.

The market reaction was mild. Nvidia stock rose about 1% on the day, which for a company of its size on a $13 billion deal is closer to a shrug than an endorsement. That is consistent with how investors have treated Nvidia's recent capital deployment: the $3.5 billion MediaTek investment drew a similar response. When quarterly revenue runs to $96.2 billion, a $13 billion acquisition is a line item rather than a bet-the-company move.

What Happens Between Now and Close

A first-half 2027 close means several quarters of regulatory review, and a deal of this size in this sector will attract attention from more than one jurisdiction. Nothing changes operationally in the meantime — Hugging Face continues to run as it does today.

For developers, the practical advice is unremarkable: nothing about your workflow changes this year, and the commitments made this week are worth holding the acquirer to next year. For anyone tracking our stock trading coverage, the more interesting signal is what it says about where Nvidia thinks the moat is. Buying the software distribution layer around your hardware is the move of a company planning for a market where the chips are no longer the only scarce thing.

Sources: NVIDIA Newsroom — September 3, 2026; CNBC — September 3, 2026; SEC Form 8-K, NVIDIA Corp — September 2, 2026.