Nvidia is buying the commons. The company that makes the chips almost every AI model is trained on has agreed to acquire Hugging Face — the neutral, open platform where the rest of the industry shares its work — for $12.93 billion.

The deal, announced today, hands Nvidia the closest thing the AI world has to a town square. More than 18 million developers, researchers and creators use Hugging Face to share over 3 million models, 500,000 datasets and a million applications; some 200,000 companies build on it. If GitHub is where the world keeps its code, Hugging Face is where it keeps its models — and it has been, pointedly, owned by no one with a horse in the race.

That is what changes. The deal is not done — it is subject to regulatory approval and NVIDIA expects it to close in the first half of 2027 — and of the headline $12.93 billion, roughly $11.9 billion goes to Hugging Face's shareholders, with up to another $1 billion set aside to retain its staff. But the direction is set.

The promise that gives the game away

Nvidia clearly understands the worry, because it led with the answer. "Hugging Face will remain an open platform for the entire AI ecosystem," the company 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. NVIDIA compute will not be required to build on or deploy through Hugging Face."

Read that last sentence again. The single most valuable property of Hugging Face is that it is neutral — you can pull a model and run it on anyone's hardware, in anyone's cloud. The moment a promise has to be made that Nvidia's chips "will not be required," you know exactly which concern the deal raises: that the industry's shared, hardware-agnostic layer is now owned by the most powerful hardware company in it.

Not a hostile takeover of an enemy

To be fair to the logic, this is not a raider buying a rival. Nvidia is already the largest single contributor of open models and data to Hugging Face — it has released more than 500 models and 250 datasets there. Jensen Huang framed the acquisition around that history, arguing that "open models let startups, businesses, universities and public institutions build on advanced capabilities without training every model from scratch," and that the plan is to scale the platform while preserving the openness that made it matter.

There is a version of this where Nvidia's money and infrastructure make the commons bigger and faster for everyone. Hugging Face has never been a cash machine; a deep-pocketed owner committed to keeping it open could be good for the ecosystem.

The question ownership doesn't answer

But contribution and control are different things, and a pledge is not a structure. Once the deal closes, Nvidia will set Hugging Face's roadmap, its defaults, its integrations and its commercial terms — the quiet levers that shape which hardware feels frictionless and which feels like swimming upstream, regardless of what is technically "required." The platform spent years earning trust precisely by not being anyone's storefront. It is now the storefront-adjacent property of a company whose chips it will be forever tempted to favour.

None of that is inevitable, and the openness commitments will be judged over years, not in a press release. Hugging Face has guarded its independence before, and its credibility is the whole asset Nvidia just paid nearly $13 billion for — which is the strongest reason to think Nvidia will tread carefully. But the structural fact stands: as of today, the open-source AI ecosystem's most important shared resource has a buyer, and that buyer is the most dominant company in AI. Whether that turns out to be stewardship or capture is now the industry's problem to watch.