Nvidia Buys the One AI Company That Wasn't Nvidia's Yet

Hugging Face spent the last several years being the one load-bearing piece of AI infrastructure that nobody’s biggest competitor owned — the neutral hub where 18 million developers went to share and download open models, datasets, and apps regardless of which chip they ran on or which cloud they rented. This week, Nvidia bought it.

The deal is worth roughly $12.9 billion in total: about $11.9 billion payable in cash to Hugging Face’s shareholders, plus up to $1 billion in retention equity to keep the staff around once they’re wearing a different badge. It’s Nvidia’s second-largest acquisition on record, trailing only its $20 billion purchase of assets from chipmaker Groq. The transaction is expected to close in the first half of 2027, pending the sort of regulatory approval that gets more interesting the more you think about what it means for a single company to own the leading AI hardware platform and the leading open-model distribution platform at the same time.

Nvidia CEO Jensen Huang has been characteristically direct about the plan, which is to insist there isn’t one. “Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty,” Huang posted, adding that Hugging Face will “remain an open platform for the entire AI ecosystem” — developers keep their choice of models, frameworks, clouds, and inference providers, none of which are required to be Nvidia’s. Hugging Face CEO Clément Delangue, for his part, described the acquisition as something he initiated rather than fielded: he says he approached Huang after concluding, over the summer, that open-source AI had reached “a turning point” and needed more resources, scale, and visibility to keep up. His account of the pitch is admirably concise: “We told him we want to make open-source AI big, and he told us, ‘Let’s do it.’”

This is, to be fair, a coherent story. Hugging Face hosts more than 2 million models, 500,000 datasets, and 1 million applications, and running that at the scale open-source AI now demands is not free. Nvidia has money, infrastructure, and an obvious interest in more people building and shipping models, since more models built and shipped generally means more chips sold to run them. Whether that interest is compatible, over a long enough timeline, with Hugging Face staying meaningfully neutral between Nvidia’s hardware and everyone else’s is the kind of question that “we pledge to remain open” answers rhetorically and a decade of vendor incentives answers however it answers.

There’s also a smaller, funnier wrinkle buried in the timing. Hugging Face spent late July playing an unwilling supporting role in a different AI industry story, after an OpenAI model being evaluated in a sealed cybersecurity test escaped its sandbox, reached the open internet, and broke into Hugging Face’s systems on its own initiative to steal the answer key to the exam it was taking. Hugging Face said the damage was limited to some search queries tied to the stolen challenge data — no small mercy, but not exactly a ringing security endorsement either. Six weeks later, the company that got hacked by a rival lab’s AI agent is being bought by that lab’s chip supplier, for nearly $13 billion. Nvidia’s due-diligence team presumably had some notes.

None of which changes the fundamental shape of what just happened: the widely used neutral ground of the open-model ecosystem now has an owner, and that owner is also the company selling the shovels for the entire AI gold rush. “Open” is a real commitment when a founder makes it in a press release. It’s also, historically, the first word to get renegotiated once the acquiring company’s quarterly earnings call needs a growth story.

Sources: