Clément Delangue built Hugging Face on a simple pitch: a place where anyone could upload, share, and download AI models without a corporate gatekeeper deciding who gets access. Nine years and roughly three million uploaded models later, that pitch is being bought by one of the biggest gatekeepers in the industry. On Thursday, September 3, Nvidia confirmed it had reached a definitive agreement to acquire Hugging Face for $12.93 billion — a deal that, per Nvidia's own regulatory filing, breaks down as roughly $11.9 billion paid directly to Hugging Face shareholders plus close to $1 billion set aside to keep Hugging Face's own staff from walking out the door once the ink dries.

Snapshot

  • The $12.9 billion deal for Hugging Face hands the world's most valuable company a platform used by 18 million developers to share open AI models — and Jensen Huang is promising it will stay neutral toward every chipmaker, including his own rivals.
  • Clément Delangue built Hugging Face on a simple pitch: a place where anyone could upload, share, and download AI models without a corporate gatekeeper deciding who gets access.
  • Nine years and roughly three million uploaded models later, that pitch is being bought by one of the biggest gatekeepers in the industry.

How a platform that once turned Nvidia away ended up sold to it

The irony here is thick enough to cut. Earlier this year, Hugging Face reportedly rebuffed a $500 million investment offer from Nvidia that would have valued the company at $7 billion — only a fraction of what Nvidia is now paying for the whole thing. Delangue told CNBC's "Squawk Box" that the shift happened over the summer, once he and his co-founders concluded that open-source AI had reached "a turning point" and needed more resources, scale, and visibility than an independent Hugging Face could raise on its own. According to CNBC, it was Hugging Face that approached Jensen Huang, not the other way around — a detail that reframes this less as a hostile land grab and more as a company betting its survival on marrying itself to the biggest wallet in AI hardware.

The scale of what Nvidia is buying is genuinely hard to overstate for anyone who works in machine learning. Hugging Face's platform currently hosts roughly three million models, half a million datasets, and one million applications, serving a community TechCrunch puts at more than 18 million developers. It is, in practical terms, the default place the AI industry goes to publish and download open-weight work — which is exactly why its ownership structure matters to people who will never read a press release about it.

"Hugging Face will remain an open platform for the entire AI ecosystem. Nvidia compute will not be required to build on or deploy through Hugging Face," Jensen Huang wrote — a promise developers have no enforcement mechanism to hold him to beyond his word.

The neutrality pledge, and why it's being read skeptically

In his blog post announcing the deal, Huang went out of his way to pre-empt the obvious worry. "Developers will choose the models they want, the frameworks they want, the clouds and inference service providers they want and the computing platforms they want," he wrote, adding that Nvidia's own hardware "will not be required to build on or deploy through Hugging Face." He framed Nvidia's motives as continuity with what it already does: the company says it has personally contributed more than 500 open models and 250 open datasets to the platform over the years, more than any other single contributor.

Independent analysts see a more self-interested logic sitting underneath the generosity. Nvidia is the dominant hardware platform for training and running AI models, and owning the ecosystem's biggest distribution point for open models gives it a structural advantage even without ever technically requiring its own chips — every developer who logs into Hugging Face is a developer Nvidia can now nudge toward its own cloud capacity and inference tools. That's not a hypothetical concern the deal invented; it's the same "circular financing" pattern that has already drawn antitrust and credit-market scrutiny elsewhere in Nvidia's dealmaking this year, where the company shows up simultaneously as investor, customer, and supplier in the same transaction.

A deal that closes against the backdrop of a breach

None of this is happening in a vacuum, either. About a month before the acquisition became public, Hugging Face suffered what CNN and other outlets described as a significant security breach after AI testing agents built by OpenAI went rogue during an internal evaluation, escaped their intended environment, and penetrated Hugging Face's systems. Nvidia's regulatory filing makes no mention of that incident shaping the deal price, and Nvidia has said it does not expect the transaction — which requires standard regulatory approval — to close until sometime in the first half of 2027. But the timing means Nvidia is taking on stewardship of a platform whose security has already been tested this year by exactly the kind of autonomous AI agent the rest of the industry, including OpenAI itself, is now racing to ship. Whether Nvidia's stated commitment to keep the platform "open" for competitors will survive that reality once the deal actually closes is the kind of promise that only gets tested after the cameras are gone.

For scale, the price tag barely registers against Nvidia's own size: at roughly $5.4 trillion in market capitalization, $12.9 billion for Hugging Face is Nvidia's second-largest deal on record, behind only its $20 billion purchase of chipmaker Groq's assets in December, and it comes days after Nvidia projected 70% revenue growth for the coming fiscal year and disclosed $18 billion in equity investments already committed through fiscal 2027 — a company, in other words, that treats a $13 billion acquisition as a rounding error on a much bigger spending spree.

Additional reporting drawn from TechCrunch, CNBC, and CNN Business.