An unsigned deal, a very signed history
Nvidia has reportedly agreed to acquire Hugging Face, the platform widely used to host and share open-source AI (artificial intelligence) models, datasets, and benchmarks, for $12.9 billion. The Information first reported the agreement on the night of August 26, 2026, citing an unnamed source, after Business Insider had reported takeover interest the previous weekend. No agreement has been signed, and neither company responded to requests for comment from the outlets that broke the story, so the deal could still change shape or fall apart entirely.
What makes the reported price notable is what came before it. Hugging Face raised a $235 million Series D round in 2023 at a $4.5 billion valuation, led by Salesforce Ventures with Alphabet's GV, IBM Ventures, and Nvidia itself among the participants. In late 2025, Nvidia came back with an offer to invest $500 million at a $7 billion valuation. Hugging Face turned it down, reportedly out of concern that a single dominant investor could end up shaping the company's direction. The figure now being reported, $12.9 billion, is roughly triple where the company sat in 2023, and nearly double the valuation Nvidia itself proposed less than a year ago.
Why a chipmaker wants a model-hosting company
Hugging Face, about a decade old and often described as a GitHub for AI models, generates roughly $150 million in annual revenue and was reported close to profitability as of July 2026. For Nvidia, owning it outright rather than holding a minority stake would reportedly serve three purposes at once.
The first is defensive: rival chipmakers are building hardware specifically aimed at running open-source models, and controlling the platform much of that ecosystem is built on gives Nvidia leverage over how those models get optimized and deployed. The second is expansion: it would push Nvidia further into the software and model layer, beyond the chips it is best known for. The third is a route back into cloud computing. Hugging Face lets developers rent compute directly on its platform, which would give Nvidia a foothold in cloud-style compute rental after it had already scaled back its own DGX Cloud offering.
The wider pattern this fits into
Nvidia is not the only infrastructure company trying to move up the stack from hardware into the software layer that sits on top of it. The same dynamic shows up across the industry: chipmakers, cloud providers, and infrastructure vendors are all trying to own more of the path between a model and the developer who deploys it, rather than compete purely on raw compute performance. Hugging Face's position, as the default place many developers go to find, share, and test open-source models, makes it an unusually direct route into that layer for whoever ends up owning it. That is also precisely why a deal like this draws more scrutiny than a typical acquisition of similar size: the platform's value to the industry rests partly on being seen as neutral ground.
Where Hugging Face's leadership stands
Hugging Face chief executive Clément Delangue has been a consistent public advocate for open-source AI models over closed alternatives, arguing that openly available models are likely to end up dominant. He and Nvidia chief executive Jensen Huang have both put their names to letters backing open-source AI development, part of a broader industry argument that keeping model weights open matters for the US to stay competitive against Chinese AI development.
That stance sits a little uneasily next to a $12.9 billion acquisition by one of the industry's largest chipmakers. An independent Hugging Face and a Hugging Face owned by Nvidia are different things to a developer community that has, so far, trusted the platform partly because it wasn't owned by any single hardware vendor.
What it means in practice
Nothing is final. Until a deal is signed, the terms reported here, the price, the structure, even whether it happens at all, should be treated as exactly that: reported, not confirmed. But the trajectory is worth watching regardless of how this specific number lands. A rejected $500 million offer becoming a reported $12.9 billion agreement inside about nine months says more about how much strategic value chipmakers now place on controlling the software layer than any single funding round would.
If it closes, the practical effects would show up gradually rather than overnight. Hugging Face's existing open-source commitments and hosted models would not disappear, but decisions about compute pricing, which cloud partners get preferential integration, and how aggressively the platform pushes Nvidia-optimized tooling would no longer be made by a fully independent company. Teams that rely on Hugging Face for hosting, model discovery, or rented compute should expect Nvidia's incentives, not just Hugging Face's, to start shaping the roadmap, and it would be worth watching whether rival chipmakers respond by investing more heavily in alternative hosting platforms rather than risk depending on a competitor's property.
