What Nvidia is reportedly buying

Nvidia has agreed to acquire Hugging Face, the open-source hub where developers share and download AI models, for roughly 12.9 billion dollars (about 11.9 billion euros or 10.2 billion pounds), The Information reported on August 26, citing a person familiar with the matter.

Neither company has confirmed a signed agreement. Business Insider and Gizmodo reported days earlier that Hugging Face, founded in 2016 and often called the GitHub of AI, was working with a bank to seek a sale near 13 billion dollars, and TechCrunch and CNBC have since reported the same figure Nvidia agreed to pay. The talks could still fall apart before a deal is formally announced.

A valuation that almost tripled in nine months

Hugging Face has moved through three very different price tags since 2023, and the jump between the last two says more about this deal than the headline number does.

DateEventValuation
2023Series funding round$4.5 billion
Late 2025Nvidia investment offer, declined by Hugging Faceabout $7 billion
August 2026Reported Nvidia acquisition (unconfirmed)about $12.9 to $13 billion

Nvidia offered to invest at a valuation near 7 billion dollars in late 2025, and Hugging Face turned it down. Less than a year later the reported price is nearly double that, the kind of jump that shows up when a buyer moves from wanting a stake to wanting full ownership.

Why the neutral hub matters more than the chips

Hugging Face's real asset is not a chip or a single model; it is the position it holds as the default distribution point that open-weight models from every AI lab flow through, Nvidia's own competitors included.

OpenAI, Google, Amazon and Anthropic are each building or backing custom silicon specifically to cut their dependence on Nvidia GPUs, and many of the open models and tools that come out of that work still get published, benchmarked and downloaded through Hugging Face. Owning the hub does not touch those chip programs directly. It does put Nvidia inside the pipeline that every rival's open-model strategy currently depends on.

What this means for an EU or UK operator

European and UK teams have generally treated Hugging Face as neutral infrastructure, a shared library any vendor's models pass through without favoring one chip maker over another.

That assumption stops holding the moment the library has an owner with a stake in the outcome. A procurement or infrastructure decision built partly on Hugging Face's neutrality, its model availability, pricing, hosting terms, dataset access, is now a decision resting on one company's incentives, and those incentives point toward Nvidia hardware.

The diligence question procurement teams should ask now

Procurement teams should start with an inventory of which production systems, fine-tuned models, or datasets pull from Hugging Face, and how easily each could move if hosting terms or access changed.

Self-hosted model registries, GitLab and GitHub package hosting, and the EU sovereign-cloud providers building their own model-hosting layers all become more attractive as hedges once the largest open hub sits inside one chip vendor's balance sheet. None of that requires migrating today. It requires knowing the exit exists before you need it.