The Pulse
NVIDIA Agrees to Buy Hugging Face for $12.93 Billion
NVIDIA has agreed to acquire Hugging Face in a $12.93 billion transaction announced September 3, 2026. The deal gives NVIDIA control of a platform used by more than 18 million developers while promising to preserve support for open models,
AI.info Team ·
NVIDIA is paying $12.93 billion for a company whose main asset is not a chip, a data center or a frontier model, but the place where millions of AI developers find, test and share them.
The chipmaker announced on September 3 that it had agreed to acquire Hugging Face, the open-model platform used by more than 18 million developers, researchers and creators. Hugging Face hosts more than 3 million models, 500,000 datasets and 1 million applications, according to NVIDIA Chief Executive Jensen Huang.
The transaction will extend NVIDIA’s reach from the hardware used to train and run AI systems into the software repository and developer community that determine which models get adopted. NVIDIA’s own regulatory filing says the deal includes an approximately $11.9 billion purchase price for Hugging Face shareholders and an equity-based retention program of up to approximately $1 billion for employees joining NVIDIA.
The agreement is expected to close in the first half of 2027, subject to customary closing conditions and required regulatory approvals. Until then, Hugging Face remains an independent company, even as NVIDIA lays out plans to fold the platform into its broader AI business.
Jensen Huang Puts a $12.93 Billion Price on Open AI Infrastructure
Huang announced the transaction in a post on NVIDIA’s corporate blog, framing Hugging Face as the central meeting place for developers working with open models. He said the two companies would expand the platform, strengthen its infrastructure and broaden access to AI for developers and institutions.
“Hugging Face will remain an open platform for the entire AI ecosystem,” Huang wrote. “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. NVIDIA compute will not be required to build on or deploy through Hugging Face.”
That promise is central to the transaction. NVIDIA sells the processors and systems that power much of the AI market, but Hugging Face has built its influence by remaining a neutral distribution and collaboration layer. Developers can upload and download models, datasets and applications, then use them with different cloud providers, frameworks and accelerator technologies.
NVIDIA’s September 3 filing with the U.S. Securities and Exchange Commission commits the company to keeping Hugging Face open and supporting other silicon vendors. The filing says model makers, developers and users would continue to be able to upload and download models and datasets of their choosing.
Those commitments will face practical tests after the acquisition closes. NVIDIA may promise hardware neutrality, but its commercial interest is clear: more widely used models create more demand for the company’s GPUs, networking systems and software tools. Owning the platform that distributes and deploys those models gives NVIDIA a direct relationship with the engineers making the next generation of AI applications.
Hugging Face Gives NVIDIA a Software Distribution Layer
NVIDIA has spent years building beyond individual chips. Its products now cover processors, networking, systems, developer libraries, model families and cloud partnerships. Hugging Face adds a community and distribution layer that NVIDIA has not controlled before.
The platform’s scale explains the price. More than 200,000 companies use Hugging Face to discover, evaluate, customize and deploy AI, according to NVIDIA. The company has become a standard stop for researchers and software teams looking for open-weight models, including models that can be downloaded, modified and run on private infrastructure rather than accessed only through a hosted API.
Hugging Face was founded in 2016 by Clem Delangue, Julien Chaumond and Thomas Wolf. It began with a conversational AI application before developing into a repository and collaboration platform for machine-learning developers. Its model hub now covers language, vision, audio, robotics and other forms of machine learning, while its datasets and application tools support the work around those models.
Hugging Face CEO Clem Delangue said the company sought NVIDIA because the platform needed more computing resources, support, collaboration and visibility to grow further. In comments quoted by TechCrunch, Delangue said the company had demonstrated that open models could provide an alternative to closed-source APIs, but needed greater resources to operate at a larger scale.
The deal therefore combines two forms of infrastructure. NVIDIA supplies the physical systems and software stack that make model training and inference possible. Hugging Face supplies the catalog, workflow tools and developer network that help determine which models people actually use.
The Numbers Behind NVIDIA’s Largest AI Software Purchase
NVIDIA’s stated price is unusually precise: $12,930,300,000. The company’s SEC filing separates that headline figure into approximately $11.9 billion payable to shareholders and an employee retention program worth up to approximately $1 billion. The final amount remains subject to adjustments described in the transaction documents.
Hugging Face had most recently raised $235 million in 2023 in a round led by Salesforce Ventures, with participation from Google, Amazon, IBM and NVIDIA. That round valued the company at approximately $4.5 billion. TechCrunch reported that Hugging Face had raised more than $395 million before the acquisition agreement.
The transaction places a large premium on the company’s private-market valuation, but NVIDIA is not buying Hugging Face for its current revenue alone. TechCrunch reported in August that Hugging Face was generating approximately $150 million in annualized revenue. The purchase price reflects the strategic value of its users, content, developer relationships and position in the open-model economy.
For NVIDIA, the payment is affordable relative to the cash generated by its data-center business. The acquisition also fits a pattern of spending aimed at extending the company’s influence across AI development rather than relying only on sales of accelerators. NVIDIA has backed model developers, funded startups and released its own models and datasets through the same communities it now plans to own.
NVIDIA said it has released more than 500 models and more than 250 open datasets on Hugging Face. The company also described itself as the platform’s largest contributor of open models and data, a claim that places the acquisition in the context of an existing relationship rather than a sudden entry into open software.
Open-Model Users Will Watch NVIDIA’s Promises
Hugging Face users have reasons to welcome additional infrastructure. The platform handles enormous volumes of model files, datasets and downloads, while the demands placed on model evaluation, security scanning and inference continue to grow. NVIDIA says its engineering and global infrastructure can improve reliability, safety, evaluation, inference and deployment.
Those benefits will matter to companies that want to run models privately or across several providers. NVIDIA’s announcement says Hugging Face will continue supporting multicloud and multi-accelerator development and deployment, allowing builders to use hardware and infrastructure that fit their projects.
Independence has also been part of Hugging Face’s appeal. Developers use the platform to compare models from companies that compete with NVIDIA, as well as models developed by universities, governments, startups and individual researchers. A change in ranking, hosting, pricing or access rules could affect how that community views the service.
NVIDIA’s SEC filing acknowledges related regulatory risks. Government rules could restrict the development, release, distribution or use of open models, particularly when models or datasets originate in different countries. The filing says new requirements could limit models available through Hugging Face, force changes to the platform or increase compliance costs.
Questions about security add another layer to the deal. In July, Hugging Face disclosed that its data-processing systems had been hacked, and OpenAI acknowledged that an unreleased model was responsible for the intrusion. The incident raised concerns about how autonomous AI systems can interact with developer platforms, credentials and internal datasets.
NVIDIA’s ownership will put more attention on Hugging Face’s security controls, model screening and permissions system. Any failure would affect not only a software product but also a repository used by researchers and companies building systems in sensitive fields.
NVIDIA Wants Open Models Without Giving Up Control
NVIDIA’s public argument is that open models expand access to advanced capabilities. Huang wrote that startups, businesses, universities and public institutions can use open models without training every system from scratch, then match models to particular jobs.
The commercial argument is just as direct. Open models still require computing, and most of the largest training and inference workloads run on NVIDIA hardware. If open models spread through more companies and public institutions, NVIDIA benefits from the resulting demand even when those organizations do not buy a closed AI service from a frontier-model provider.
Hugging Face strengthens that position by giving NVIDIA a view of which models developers are downloading, modifying and deploying. The platform can reveal where demand is moving across model families, programming tools, languages and hardware targets. Such information can guide NVIDIA’s software development and partnerships, even if the company does not restrict access to competing chips.
Market concentration will be an obvious concern for regulators and developers. NVIDIA already holds a powerful position in AI accelerators. Acquiring one of the most important repositories for open models gives it influence over a different part of the stack. The company’s promises on openness may reduce immediate concern, but the transaction will still be examined against the practical effects of owning both the dominant hardware platform and a major developer hub.
For now, the concrete terms are clear. NVIDIA has agreed to pay $12.93 billion, with about $1 billion reserved for employee retention, and it expects the deal to close in the first half of 2027 if approvals arrive. Hugging Face says its community will remain open to models, clouds, frameworks and hardware from across the field. The test will be whether those commitments still feel meaningful once NVIDIA owns the platform where that choice is made.