Nvidia Buys Hugging Face for $12.9B, Vows to Keep It Open

What happened

Nvidia has agreed to acquire Hugging Face, the largest hub for open-source and open-weight AI models, for $12,930,300,000 — a figure that turned out to be a deliberate nod to Hugging Face's 🤗 emoji, since 129,303 is the decimal value of that emoji's Unicode code point 108. The deal was confirmed Thursday, September 3, in a blog post from Nvidia CEO Jensen Huang, ending weeks of reports that a sale was in the works 715. According to Nvidia's SEC filing, the agreement was signed September 2 and breaks down into roughly $11.9 billion for Hugging Face shareholders plus up to $1 billion in equity retention awards for employees joining Nvidia, with closing expected in the first half of 2027 pending regulatory approval 7820.

Hugging Face began in 2016 as a teen-oriented chatbot app built by French entrepreneurs in New York before pivoting into what became the de facto GitHub of machine learning 912. Nvidia says the platform now hosts more than 3 million models, 500,000 datasets and 1 million applications, serving over 18 million developers and 200,000-plus companies 15816. Hugging Face had raised about $395 million to date, most recently a $235 million round in 2023 that valued it at $4.5 billion with backing from Salesforce, Google, Amazon, Nvidia, Intel and IBM 1517. Multiple outlets note that Hugging Face reportedly turned down a $500 million Nvidia offer previously and was said to be generating around $150 million in annualized revenue this year — meaning Nvidia is paying roughly 86 times that figure 151617.

Why Nvidia wants it

Huang framed the purchase as a way to scale Hugging Face's infrastructure and widen AI access, writing that the two companies would work "to make AI more open, more capable and more accessible" 207. Hugging Face CEO Clément Delangue said the deal reflects an inflection point for open-source AI, thanking the community for proving open models can rival closed APIs, and set a goal of reaching 100 million AI builders with Nvidia's backing 199.

Analysts and outlets read strategic motive into the price tag. Fortune and other coverage point out that major Nvidia customers — OpenAI, Microsoft, Amazon and Meta among them — are building their own chips to reduce GPU dependence, making open-weight models a hedge for Nvidia's future revenue: the more developers build and deploy open models, the more compute gets sold, much of it Nvidia's 9. The New York Times quotes Forrester's Jeff Pollard describing Hugging Face as the exchange point for a large share of the AI ecosystem's models, datasets and code, giving Nvidia a closer vantage on the entire AI supply chain 20. TechCrunch adds that Nvidia stands to benefit by packaging unused compute capacity with Hugging Face's offerings 15.

Central to the pitch is a neutrality pledge. Huang stated plainly that developers will keep choosing their own models, frameworks, clouds and computing platforms, and that "Nvidia compute will not be required to build on or deploy through Hugging Face" 15187. That promise is repeated nearly verbatim across TechCrunch, Phoronix, CryptoTimes and Finelo, suggesting Nvidia is treating it as a load-bearing commitment rather than a throwaway line 1518816.

Where the reporting agrees

Across outlets — Business Insider, TechCrunch, Wired, the New York Times, Fortune, Mashable, CryptoTimes, Phoronix, Finelo, Sacra and X-based coverage of Huang's own posts — there is no real dispute about the headline facts: the price is $12,930,300,000, announced September 3, structured as roughly $11.9 billion for shareholders and up to $1 billion in employee retention equity, with closing targeted for the first half of 2027 781520. Every account also agrees on the emoji easter egg embedded in the price 109. There's consistent agreement that Hugging Face was founded in 2016, pivoted from a chatbot into an open-source ML platform, and had previously been valued at $4.5 billion in a 2023 round backed partly by Nvidia itself 91517. And nearly every source repeats Nvidia's neutrality pledge nearly word for word, indicating outlets are drawing from the same Huang blog post and press briefing rather than independent verification 1518168.

Where it doesn't

The platform's scale numbers vary depending on when and how each source counts. Nvidia's own figures, echoed by TechCrunch, CryptoTimes and Finelo, put the Hub at 3 million models, 500,000 datasets, 1 million applications and 18 million developers 15816. But a Contrary research profile cites 2.4 million models and 730,000 datasets as of January 2026, while Sacra describes 2 million-plus models and 13 million users across 500,000 organizations, and a separate company profile from swfte.com claims just over 1 million models and 250,000 datasets with only 250 employees 111713. These aren't necessarily contradictions so much as a moving target — Hugging Face's inventory appears to have grown quickly in the months before the deal — but the spread is wide enough that readers should treat any single snapshot as approximate rather than precise.

There's also a genuine reporting gap around motive and history. Fortune is the only outlet to report that Hugging Face rejected a prior $500 million Nvidia offer, attributing that detail to the Financial Times, and it notes CEO Delangue declined to confirm the figure directly, saying prior reporting on the company's fundraising "can be quite far from reality" 915. TechCrunch also references the rejected $500 million bid, but neither Nvidia nor Hugging Face has independently verified the number, meaning it rests on secondhand sourcing rather than confirmed disclosure 15.

Framing diverges more than facts do. Wired and the New York Times treat the acquisition primarily as a bet on open-source AI's future and a statement in the broader open-versus-closed model debate 1920. CryptoTimes takes a narrower financial-infrastructure angle, tying the deal to decentralized-AI narratives around projects like Bittensor and asking whether a centralized platform under Nvidia's ownership undercuts the goals of distributed AI development 8. Fortune is the only outlet to raise the compensation question directly, quoting Dolphin model creator Eric Hartford's complaint that Hugging Face never built a way for individual model creators to monetize their work — a critique that sits uncomfortably next to Nvidia and Hugging Face both profiting handsomely from a platform built largely on free community contributions 9. That's a detail no other source raises, and it points to a tension the corporate announcements do not address.

The unresolved question

The evidence across outlets does not yet settle whether Nvidia's neutrality pledge will hold in practice. Every company statement and nearly every outlet repeats the same assurance — that Hugging Face will remain open and that Nvidia compute won't be mandatory — but that is a promise made by the acquirer, not a structural guarantee, and Fortune notes that some in the community already suspect a hardware bias that the deal is likely to deepen 9. Given that Nvidia's core business depends on GPU demand, and that Hugging Face is where a large share of AI developers actually decide what to build and run, the more plausible reading is that Nvidia's commitment to openness will be tested gradually — through search rankings, default inference options, and how quickly non-Nvidia hardware backends get supported — rather than through any dramatic reversal. The reporting agrees on what was promised; it does not yet have the evidence to say whether it will be kept.