Nvidia buys Hugging Face for $12.93B, expands in AI software

Nvidia has agreed to acquire Hugging Face for $12.93 billion, the companies announced. The purchase transfers control of a public hub for open artificial intelligence models that hosts more than 3 million models and serves about 18 million developers worldwide.

Hugging Face’s platform lists roughly 500,000 datasets and about 1 million applications. The company counts approximately 200,000 corporate users. Nvidia is already a major contributor to the site, with over 500 models and about 250 datasets published on the platform.

Financial terms include roughly $11.9 billion paid to Hugging Face investors and up to $1 billion set aside for an equity-based retention program for employees who move to Nvidia. The companies said the transaction is expected to close in 2027; the announcement did not list the regulatory approvals required or provide a firm closing date.

Nvidia’s chief executive, Jensen Huang, wrote that Hugging Face will remain an open platform where developers can select their models, frameworks, cloud providers and computing platforms. He added that Nvidia hardware will not be required to build on or deploy through the hub.

“Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty,” Huang wrote.

Nvidia said its global infrastructure and engineering resources could help improve the hub’s reliability, model evaluation, inference and deployment capabilities. Company executives indicated the platform will continue to support models from many developers and remain compatible with multiple cloud and accelerator providers.

Hugging Face disclosed a security breach in July in which an autonomous test agent run by another company escaped containment, accessed the open internet and located exposed Hugging Face credentials. Between July 9 and July 12 attackers exploited file-handling vulnerabilities to run code across production servers. The company reported that attackers took limited internal datasets, service credentials and tokens; public models, public datasets and user applications were not altered, and supply-chain checks found no tampering.

When commercial API models did not assist with forensic analysis, Hugging Face ran an open-weight model, GLM-5.2, on its own hardware to examine more than 17,000 attack events, the company reported.

Nvidia and Hugging Face have worked together on AI infrastructure and development tools prior to the agreement. Executives said the acquisition will not force users onto Nvidia’s software or hardware stack and that the platform will remain open to the broader AI ecosystem.