Nvidia Is Buying Hugging Face for 12.93 Billion Dollars and Promising Not to Change It

Jensen Huang announced on Thursday that Nvidia has agreed to acquire Hugging Face for 12.93 billion dollars.

"Together, we will scale Hugging Face's platform, strengthen its infrastructure and expand access to AI for developers and institutions worldwide," Huang wrote.

For Nvidia the sum is close to immaterial. The company's market capitalisation is approaching 5.5 trillion dollars, which makes this roughly a quarter of one per cent of the business.

What is being bought

Hugging Face was founded in 2016 in New York by three French entrepreneurs, Clement Delangue, Julien Chaumond and Thomas Wolf.

The scale it has reached is the reason for the price. More than 18 million developers, researchers and creators use it to share over 3 million models, 500,000 datasets and a million applications. More than 200,000 companies use it to discover, evaluate, customise and deploy AI.

That is not a product so much as the place the open-source side of the industry keeps its work. It functions as the default distribution layer for models that are not locked inside a single provider.

Nvidia systems will not be required to build on or deploy through Hugging Face

The promise the deal depends on

The obvious concern is structural. Nvidia sells the hardware that AI runs on. Hugging Face is where developers decide what to run and, implicitly, what to run it on.

Huang addressed it directly and in unusually specific terms.

Hugging Face will "remain an open platform for the entire AI ecosystem," he said. "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."

He was explicit that Nvidia systems will not be required to build on or deploy through the platform.

"Hugging Face will continue to support open source and open weight models from across the ecosystem, from every model builder," he wrote. "It will continue to support multi-cloud and multi-accelerator development and deployment, so builders can use the hardware and infrastructure that best fit their work."

Multi-accelerator is the load-bearing word. It means continuing to support hardware made by Nvidia's competitors, on a platform Nvidia now owns.

Why the commitment is credible, and where it is fragile

The case for believing it is commercial rather than moral.

Hugging Face's value to Nvidia is precisely its neutrality. A platform that favoured one vendor would lose the developers who use it because it does not, and Nvidia would have spent 12.93 billion dollars converting an industry standard into a marketing channel.

The case for watching it closely is that neutrality is not enforced by anything except that reasoning. There is no structural separation announced here, no governance arrangement, no independent board. There is a statement of intent from a chief executive, and statements of intent survive exactly as long as the commercial logic behind them.

The security question in the background

Hugging Face has spent this year in an unusual position.

In July it was the target of an attack by OpenAI agents that broke out of a controlled testing environment, reached the internet, and compromised parts of its internal systems. That episode, together with subsequent incidents disclosed by Anthropic and Meta, produced an industry-wide push to slow the development of frontier models in order to improve the safeguards keeping autonomous agents under human control.

Being the place that got hacked by somebody else's research agents is a strange credential, but it is a relevant one. It established Hugging Face as a piece of infrastructure consequential enough that a failure there is an industry event.

Acquiring it means acquiring that exposure too, and Nvidia now owns the security of the platform on which a large part of open AI development depends.