Nvidia is closing in on a $12.9 billion acquisition of Hugging Face, according to reports that surfaced on Aug. 27 in CNBC, Bloomberg, and Fortune, following TechCrunch’s initial story a day earlier citing The Information. An anonymous person familiar with the matter told CNBC the “confirm acquisition [by Nvidia] has been part of ongoing and recent talks,” though nothing has been signed and The Next Web noted on Aug. 29 that any deal remains subject to regulatory approval.

The price is remarkable less for its size than for its ratio. Hugging Face was valued at $4.5 billion in a 2023 round of $235 million led by Salesforce Ventures, with Alphabet’s GV and IBM Ventures participating. Late last year, Clément Delangue reportedly turned down a $500 million investment from Nvidia at a $7 billion valuation. Revenue has since climbed to roughly $150 million annually, up from about $100 million just two months earlier, with Delangue telling TechCrunch the company is “close to profitability.” A $13 billion price against $150 million puts the multiple near 86 times revenue.

Hugging Face hosts more than three million open-source models used by roughly 500,000 organizations, and its Inference Endpoints span pricing from about $0.03-per-hour CPU instances to $80-per-hour eight-GPU H100 clusters. It’s, functionally, the neutral distribution layer for open weights from OpenAI, Google, Amazon, Anthropic, and Chinese labs like Moonshot AI, whose Kimi K3 sits alongside Western releases. TechCrunch’s read is direct: the platform “gives customers more alternatives to those closed labs, which in turn keeps more of the market dependent on Nvidia’s hardware.”

That’s the strategic logic. The neutral commons stays neutral only until someone owns it.

Nvidia disclosed $96.2 billion in quarterly revenue the same day the deal leaked, with a 70 percent revenue jump forecast for next fiscal year and shares up 4 percent in after-hours trading. Delangue and Jensen Huang co-signed a letter to Washington with more than 20 companies urging support for open models, and their joint Training Cluster as a Service already ties the two firms operationally. Third-party inference providers like Together AI, SambaNova, and Groq compete on the platform today; whether they still compete on equal footing under Nvidia’s ownership is the antitrust question.

The historical rhyme is YouTube’s 2006 sale to Google. A neutral platform for user-generated content became a distribution arm of a search monopoly, and the terms of neutrality quietly changed over the following decade. Delangue has called China “clearly dominating” on open source. Stripe reportedly paid more than $7 billion for OpenRouter, valued at $1.3 billion during its Series B in May. The consolidation is priced in. What’s being negotiated now is who owns the substrate the rest of the AI economy runs on.

Sources