New reports outline deal talks between the two companies as open-source AI technology comes more into focus

Hugging Face, co-founded by Thomas Wolf, was valued at $4.5 billion through a 2023 funding round.

As open-source artificial intelligence gains momentum, Nvidia may be making a sizable acquisition.

The Information reported late Wednesday that the company had struck a $12.9 billion deal to acquire Hugging Face, a repository for open-source AI tools. Business Insider reported earlier in the evening that the parties had been holding talks over a deal that could value Hugging Face at north of $13 billion.

Nvidia (NVDA) and Hugging Face didn't immediately respond to MarketWatch's requests for comment.

Open-source AI refers to models that can be freely used and modified. This contrasts with closed-source AI, which is designed to be controlled and private.

"The world will need both closed models and open models," Nvidia CEO Jensen Huang noted on the company's Wednesday earnings call. "And both closed models and open models are skyrocketing in use."

Huang added that "nearly all open models run on Nvidia."

Hugging Face positions itself as a platform for AI developers. The company says it provides easy access to the sorts of tools that AI builders need to create their own AI applications and collaborate with others in the technology world. Hugging Face houses open-source models, data sets and applications.

While users can browse, download content and conduct other activities for free on the Hugging Face platform, the company also offers a Pro subscription product that it says lets people "explore the entire spectrum of AI, covering text, image, video, audio, 3D and more." The subscription also affords access to greater computing power, or "up to 40 minutes of RTX Pro 6000 Blackwell," which is a reference to a type of Nvidia hardware.

The company said that most of its employees work remotely, though it has offices in Brooklyn and Paris.

Open-source AI is heating up as developers look for more cost-efficient ways to build with AI. By publishing things like model weights, open-source players make it so others can easily build on what they've created.

"Increasingly the world is running on open and local models because they're more affordable than the big APIs and private by design," Hugging Face CEO Clément Delangue said in a July press release, referring to application programming interfaces.

Delangue said on CNBC earlier this month that he thought China was already beating the U.S. when it came to open-weight models. He argued that U.S. companies have been "building in silos" while Chinese players have been more open to collaboration.

Don't miss: Meet Kimi K3, the newest Chinese AI model haunting Silicon Valley

Hugging Face gained attention earlier this summer when OpenAI models breached controls and were able to compromise the company's systems. This offered a glimpse of the potential for AI models to go rogue, as the models were meant to be isolated from the internet.

Through a 2023 funding round, Hugging Face fetched a private-market valuation of $4.5 billion.

-Emily Bary