Nvidia to buy Hugging Face for $12.9 billion, uniting GPUs with open AI models

Nvidia has agreed to buy Hugging Face for $12.93 billion, the chipmaker confirmed Wednesday, ending months of speculation about a deal joining the dominant maker of AI hardware with the largest hub for open-source AI models. Hugging Face’s platform hosts three million models, one million applications, and half a million datasets, used by more than 18 million developers.

In a blog post announcing the deal, Nvidia chief executive Jensen Huang said Hugging Face would keep operating as an open platform. “Developers will choose the models they want, the frameworks they want, the clouds and inference service providers they want and the computing platforms they want,” Huang wrote, adding that Nvidia compute would not be required to build on or deploy through Hugging Face. He noted that Nvidia has already published more than 500 models and 250 open datasets on the platform. Owning the hub where much of the open-model world gathers also gives Nvidia more influence over the tools built for its chips, and, as TechCrunch has reported separately, a channel to sell spare data center capacity to enterprise customers bundled with Hugging Face’s offerings.

Hugging Face, founded in 2016, has raised more than $395 million to date, according to Crunchbase, with its last round in 2023: $235 million led by Salesforce Ventures, joined by Google, Amazon, IBM and Nvidia. The company turned down a $500 million offer from Nvidia last year, the Financial Times has reported, and was recently said to be generating $150 million in annualized revenue, according to The Information. In July, Hugging Face co-founder and chief executive Clem Delangue said the company’s growth was pushing it “close to profitability.”

Delangue framed the sale on X as the way to scale what Hugging Face had already proven. “But for it to happen at [a] larger scale, it needs more compute, more support, more collaboration, and more visibility. That’s why we went to talk to Jensen, who offered to do exactly that with us,” he wrote. Huang has separately argued that open-weight models, whose underlying parameters are published for anyone to inspect or modify, matter beyond developer convenience: he has co-signed a letter urging U.S. support for open models to compete with China, and told analysts on a recent earnings call that open models underpin a new wave of autonomous cybersecurity systems. Delangue said in July that one of Nvidia’s open models helped Hugging Face fend off a cyberattack after a proprietary model failed to stop it, days after OpenAI acknowledged that one of its own unreleased models had breached Hugging Face’s systems.

What happens next is whether Hugging Face’s promised independence survives contact with its new owner. Huang has committed to keeping rival hardware, cloud and inference providers on the platform, but Nvidia has not disclosed how that commitment will be enforced or for how long it will hold.