New York (AFP) – American semiconductor giant Nvidia will take control of the Franco-American artificial intelligence (AI) platform Hugging Face for $12.9 billion.

Published on: 09/03/2026 - 3:17 p.m. Modified on: 09/03/2026 - 3:40 p.m.

The operation, reported at the end of August by the website The Information, is a new illustration of the diversification of the Santa Clara (California) group, which is positioning itself, through participations, acquisitions or guarantees, throughout the AI ​​ecosystem.

“I am honored that Clem (the co-founder and boss Clément Delangue) came to see me as we approached the next chapter and thought that Nvidia could be a great home for the company,” explained Jensen Huang, the boss of Nvidia, in a letter published on his group’s website.

The buyout price represents a jump compared to the valuation of 7 billion dollars which served as the basis for first negotiations with Nvidia, at the end of 2025, mentioned by the Financial Times.

The Californian company then proposed to inject $500 million into the capital of this site which offers access to a very wide range of AI models.

Created in 2016 in New York by three French people, Hugging Face, whose name is inspired by an emoji, presents itself as “the AI ​​community that builds the future”.

It is intended to be a vector for popularizing so-called open models ("open-source" or "open-weight"), that is to say downloadable and modifiable, as opposed to closed models like those of OpenAI or Anthropic.

Hugging Face offers the world's largest library of freely accessible open models. Its revenue comes from hosting services, computing power and software interfaces to manage the use of AI.

The emergence of AI agents, programs capable of carrying out tasks autonomously, evaluating the result and making modifications, has increased the need for computing power and the associated costs.

To avoid blowing up their bills, more and more companies and IT developers are turning to open models, which are often less expensive than closed AI.

After the United States lost interest for a while in these open models, the sector has once again accelerated in this niche in recent months, in part to offer an alternative to Chinese AI, essentially open and very efficient.

These factors have significantly increased the attendance of Hugging Face, valued at $4.5 billion during its last funding round, in 2023.

“Pivotal moment”

The site gained notoriety in July after the revelation of the intrusion into its systems of two OpenAI models during a test phase.

The two AIs escaped from their confined environment to access the internet and travel behind the scenes of Hugging Face to find answers to questions submitted by OpenAI developers.

The affair has heightened fears around model control, especially since similar incidents have been reported at Anthropic and China's Moonshot AI.

Generating tens of billions of cash each quarter, Nvidia has chosen to invest it in technological players, particularly start-ups, to support the development of AI.

The company intends in particular to encourage the development of an open AI ecosystem in the United States and has developed its own open access model, Nemotron.

“Over the years, (the three co-founders) Clem (Clément Delangue), Julien (Chaumond) and Thomas (Wolf) have built something remarkable: a vibrant place for the community of open model developers,” said Jensen Huang.

The participation of the three entrepreneurs in the company is not public.

The manager assured that Hugging Face "(would) remain an open platform for the entire AI ecosystem", thus responding to the fears of some of seeing the site lose part of its independence and its DNA.

“+open-source+ is at a pivotal moment,” argued Clément Delangue on X. “Thanks to the community (of developers), we have shown that it can be a complement, or even an alternative, to closed models.”

“But for the movement to gain momentum,” he continued, “you need more computing power, more support, collaboration and visibility. That's why we went to talk to Jensen, who proposed all of this.”

© 2026 AFP

The content you are trying to access does not exist or is no longer available.