The Nvidia Hugging Face acquisition is now confirmed in black and white: the two companies signed a definitive agreement on 2 September 2026, with Nvidia agreeing to pay approximately $11.9 billion to Hugging Face stockholders, plus an equity-based retention programme of up to approximately $1.0 billion for employees who join Nvidia. That brings the total announced figure to roughly $12.9 billion, which is how the deal was first reported by The Information. The transaction is expected to close in the first half of 2027, subject to regulatory approvals.
So yes, a signed deal exists. The ‘talks that could still atomize’ framing from Business Insider is already history.
The Logic Behind the Nvidia Hugging Face Acquisition
Hugging Face is where the open-source AI ecosystem lives. Developers share, download, and run models there; the platform hosts over one million repositories and serves more than 50,000 organisations. For Nvidia, buying it is less about the $150 million in annual revenue Hugging Face was recently generating (up from roughly $100 million just two months prior, per The Information) and more about architecture: who controls the rails that open-source AI runs on.
The strategic anxiety behind the deal is straightforward. OpenAI, Google, Amazon, and Anthropic are all building their own chips. Every GPU sold to an in-house silicon team is one Nvidia doesn’t sell. A thriving open-source ecosystem, one that Nvidia effectively curates and hosts, keeps a large portion of the market dependent on its hardware rather than migrating toward vertically integrated alternatives.
Nvidia has already been spending heavily to make that case. CEO Jensen Huang and 24 other signatories, including Hugging Face CEO Clem Delangue, co-signed a letter urging the US government to back open AI models rather than restrict them. Huang has since pledged to ‘scale Hugging Face’s platform, strengthen its infrastructure and expand access to AI,’ and committed that ‘Hugging Face will remain an open platform for the entire AI ecosystem,’ according to Galaxy Research.
That last bit matters. The SEC filing contains an explicit commitment: Nvidia must keep Hugging Face’s platform open, permitting model makers, developers, and users to upload and download models and datasets of their choosing, and must continue supporting other silicon vendors. In other words, Nvidia has contractually agreed not to turn Hugging Face into an Nvidia-only shop. Whether that holds up in practice five years from now is a different question.
Hugging Face Had Other Offers on the Table
Hugging Face was not short of suitors. Galaxy Research reports the company entertained bids from Salesforce and Microsoft before settling on Nvidia. That context makes the final price look less like a negotiating win for Nvidia and more like the cost of outbidding a competitive field.
The $12.9 billion figure is also a sharp jump from prior reference points. Hugging Face raised $235 million in its 2023 Series D round at a $4.5 billion valuation. That round drew a broader investor group than initially reported: alongside Salesforce Ventures, Alphabet’s GV, IBM Ventures, and Nvidia, the round also included Google, Amazon, Intel, AMD, Qualcomm, and Sound Ventures, according to TechCrunch’s coverage of the round, with total funding reaching $395.2 million at that point.
Nvidia itself tried to invest $500 million in Hugging Face late last year at a $7 billion valuation, per the Financial Times. Hugging Face declined, saying it didn’t want a dominant investor who could sway its decisions. The gap between ‘we don’t want one big backer’ and ‘we’ll accept a full buyout at $12.9 billion’ is mostly explained by arithmetic: Delangue told TechCrunch last month the company was getting ‘close to profitability,’ but near-profitability on $150 million in revenue makes a near-$13 billion price hard to walk away from.
There is also a cloud computing angle. Nvidia reportedly scaled back its DGX Cloud business roughly a year ago. Hugging Face already lets developers run models on rented compute, so owning it hands Nvidia a route back into that market. It would also give Nvidia a ready buyer for any unused cloud capacity it ends up holding under commitments made to customers.
The deal fits into a broader pattern of consolidation eating into AI infrastructure startups. Stripe agreed to acquire OpenRouter for more than $7 billion, reported on 16 August 2026, a deal that valued the company at more than five times its $1.3 billion Series B valuation from May 2026. OpenRouter, which its CEO described as the Stripe for AI, had 8 million global users and access to more than 400 models at the time of its $113 million Series B. The pattern: acquirers are paying enormous multiples for platform-layer businesses that sit between users and the underlying models.
Regulatory approval is the remaining variable. The deal won’t close until the first half of 2027 at the earliest, and antitrust scrutiny of large AI acquisitions has been intensifying on both sides of the Atlantic. An open-platform commitment baked into the SEC filing might soften that scrutiny. Then again, regulators have seen vendor commitments before.
