Nvidia to buy open-source AI hub Hugging Face for $12.9bn

Nvidia has announced its acquisition of Hugging Face, the platform that has become the default home for open-source artificial intelligence development, for $12.93bn (€11.1bn), the chipmaker's chief executive Jensen Huang announced on Wednesday. ADVERTISEMENT ADVERTISEMENT The deal brings together two companies that sit at opposite ends of the AI supply chain but have grown increasingly dependent on each other. Nvidia makes the graphics processing units that train and run almost every major AI model in use today, giving it enormous leverage over the hardware side of the industry. Hugging Face, by contrast, has built its influence by staying open and largely neutral — it hosts more than three million AI models, 500,000 datasets and one million applications, used by more than 18 million developers and 200,000 companies to build, test and deploy AI systems regardless of which company's chips or cloud they run on. For Nvidia, buying that platform means owning not just the infrastructure AI runs on, but a direct relationship with the developers who decide what gets built and how. For Hugging Face, it means access to the capital and engineering resources of the world's most valuable chipmaker at a moment when demand for computing power, and scrutiny of who controls it, has never been higher. Huang said in a blog post that Hugging Face would continue to operate as an open platform, with developers free to use whichever models, cloud providers or computing hardware they choose. "NVIDIA compute will not be required to build on or deploy through Hugging Face," he wrote, adding that the platform would keep supporting open-weight models and multi-cloud deployment "from every model builder." Hugging Face's co-founder and chief executive Clément Delangue told CNBC that he approached Huang about a deal over the summer, saying the company had reached a point where open-source AI "needed more resources, more scale, more visibility." Nvidia already held a minority stake in Hugging Face, having invested $235 million (€202m) in the company's 2023 funding round, when Hugging Face was valued at $4.5bn (€3.87bn) — meaning Wednesday's price represents close to a threefold increase in that valuation in three years. The acquisition is Nvidia's second-largest to date, behind only its roughly $20bn (€17.2bn) purchase of assets from rival chipmaker Groq late last year, and well ahead of its previous largest deal, the near-$7bn (€6bn) purchase of Israeli chipmaker Mellanox in 2019. The timing reflects broader pressures on both companies. Nvidia's dominance in AI chips has faced growing competition as OpenAI, Google and Anthropic each develop their own custom silicon to reduce reliance on Nvidia hardware; owning the platform millions of developers use daily gives Nvidia a foothold that is harder to route around. Huang has also spent months publicly campaigning for the continued spread of open-weight AI models, co-signing a letter with more than twenty other companies, including Microsoft and Meta, arguing that open models "broaden access to AI" and prevent its benefits being concentrated among a handful of firms. Delangue has made similar arguments, warning in earlier interviews that Chinese labs such as Moonshot AI are closing the gap with leading US models at a fraction of the cost, a dynamic that has fed concern in Washington over AI competitiveness. The deal also marks a return to cloud computing for Nvidia, which scaled back its own cloud unit roughly a year ago. Hugging Face already provides developers with rented computing power to run models, giving Nvidia a route back into that business without building it from scratch. Neither company has disclosed a timeline for when the acquisition is expected to close.