Nvidia has agreed to acquire Hugging Face for $12.9 billion, The Information reported Wednesday evening, citing a person familiar with the negotiations. Business Insider, which first reported the takeover interest over the weekend, put the valuation above $13 billion. It also cautioned that nothing has been signed and that the talks could still fall apart.
Neither company has commented publicly. TechCrunch noted that Nvidia's silence is itself unusual. The chipmaker normally moves fast to knock down reports it considers inaccurate.
If the deal closes, Nvidia would own the default distribution layer of open-source AI. Hugging Face, founded in 2016, is where most developers go to publish and download open models. That position is worth considerably more to Nvidia than the startup's revenue would suggest.
A Defensive Bet on Open Models
The strategic logic starts with Nvidia's own customer list. OpenAI, Google, Amazon and Anthropic are each building custom accelerators. Every one of those programs chips away at Nvidia's pricing power.
A thriving open-model ecosystem cuts the other way. It gives enterprises a credible alternative to renting intelligence from a handful of closed labs. Those enterprises still have to run the weights somewhere, and that somewhere is overwhelmingly Nvidia silicon. The company has already poured tens of billions into building open models of its own, including the Nemotron family it now serves through partner clouds.
Hugging Face chief executive Clem Delangue has spent much of the year arguing the same side. He co-signed a letter with Nvidia CEO Jensen Huang and 24 other companies urging Washington to support open-weight models rather than restrict them. Appearing on CBS's "Face the Nation" earlier this month, he said Hugging Face had used an Nvidia-modified version of a Chinese open model to defend itself after a cyberattack. He has also warned that China is "clearly dominating" open-source AI, pointing to systems such as Moonshot AI's Kimi K3 that match leading U.S. models at a fraction of the running cost.
A Back Door Into Cloud
The acquisition would double as a quiet re-entry into cloud computing. Nvidia scaled back its own DGX Cloud effort roughly a year ago. Hugging Face already rents computing power to developers running models on its platform, which would let Nvidia return to that market without rebuilding it.
There is a balance-sheet motive as well. Nvidia has committed to helping cover tens of billions of dollars in cloud deals signed by its customers. If those customers do not consume what they reserved, the leftover capacity becomes Nvidia's problem. A marketplace full of developers is a convenient place to resell it.
A Steep Multiple
The price is the hardest part to justify on fundamentals. Hugging Face raised $235 million in 2023 at a $4.5 billion valuation, in a round led by Salesforce Ventures that also drew Alphabet's GV, IBM Ventures and Nvidia itself. The Information reported the company recently generated about $150 million a year in revenue, up from roughly $100 million two months earlier. Delangue told TechCrunch last month that growth had brought it close to profitability.
At $12.9 billion, that implies a multiple somewhere north of 80 times revenue. Nvidia would not be buying the income statement. It would be buying the developer graph attached to it.
Hugging Face has said no to Nvidia before. It turned down a $500 million investment at a $7 billion valuation late last year, according to the Financial Times, saying it did not want a dominant shareholder able to sway its decisions. An outright sale sidesteps that objection, since control changes hands cleanly instead of partially.
Consolidation Season
The timing fits a broader pattern. Stripe recently agreed to pay more than $7 billion for OpenRouter, a model-routing startup valued at just $1.3 billion in May. Independent middleware layers in AI are being absorbed by whoever controls the money or the metal.
The open question is neutrality. Much of Hugging Face's value comes from being hardware-agnostic, a switchboard rather than a storefront. A vendor-owned hub may find that harder to sustain, and rival chipmakers will be watching closely to see how model support gets prioritized after any close.
