Nvidia Agrees to Buy Hugging Face for $13B

Nvidia NVDA 1.17%increase; green up pointing triangle has agreed to buy New York-based AI startup Hugging Face in a deal worth roughly $13 billion, the latest acquisition by the chip giant at the center of the artificial-intelligence boom.

The deal accelerates Nvidia’s push to promote open-weight AI models and services that challenge both proprietary models from big U.S. AI companies and open models developed in China.

Unlike proprietary models developed by companies such as OpenAI and Anthropic, open-weight models are public, enabling users to download, customize and run them on their own or third-party hardware.

Founded a decade ago, Hugging Face has grown into a leading platform for accessing open-weight AI. Engineers use it to collaborate on machine-learning projects, from model development to applications. It was the target of a recent unexpected hack by a new OpenAI model that went rogue.

Nvidia Chief Executive Jensen Huang said Hugging Face would remain an open platform. “Open weights broaden access to AI and help ensure that AI leadership is distributed across companies, institutions and communities,” he said in a statement Thursday.

“Open models let startups, businesses, universities and public institutions build on advanced capabilities without training every model from scratch,” the statement said.

Hugging Face approached Nvidia to discuss a possible deal earlier in the summer, according to Clément Delangue, the startup’s chief executive.

“I think we realized that Hugging Face and open source AI in general was at the turning point, and that it needed more, more resources, more scale, more visibility,” Delangue told CNBC’s Squawk Box Thursday.

Delangue said he approached Nvidia because it was “a perfect home” for his company.

The company was founded in New York in 2016 by three French entrepreneurs—Julien Chaumond, Thomas Wolf and Delangue—with the aim of building a chatbot for teenagers. The technology underpinning the original product—particularly tools for natural-language processing—attracted much more interest from developers than the consumer app itself.

A focus of the business became its Hugging Face Hub, which allows developers to upload and download AI models and AI data sets in much the same way software developers use GitHub to share code.

Nvidia has been at the center of a flurry of recent dealmaking activity, using its financial heft to backstop data-center financing and other deals with customers. Last month, Nvidia announced plans to join with top Wall Street investment firms to provide partial guarantees for up to $500 billion in data-center financing and soon after agreed to backstop a massive Ohio data-center project by OpenAI.

Some industry observers and executives are concerned that a lack of focus on open-source models in the U.S. will prompt businesses and countries to embrace Chinese open-weight models instead. Those models in some cases can rival the performance of U.S. models, while being cheaper to operate.

Some U.S. officials have considered restricting access to Chinese open-weight tools because of security concerns, but Huang, Mark Zuckerberg and other Silicon Valley executives have pushed back on such restrictions.

“Don’t regulate hypothetical, theoretical harm. Regulate actual and pragmatic harm,” Huang told officials from the G-20 countries this week.

Late last month, Nvidia struck a $6 billion deal with American AI startup Poolside to license its open-weight models and hire the bulk of its engineers, The Wall Street Journal reported.

Nvidia launched the Nemotron Coalition, an association of leading open-model developers including Mistral, Thinking Machines Lab and Perplexity, earlier this year and said it would share data, expertise and computing resources.

More than 18 million developers, researchers and creators use Hugging Face to share more than 3 million models, 500,000 data sets and 1 million applications, according to Nvidia.