Nvidia’s $12.9 billion Hugging Face deal is not yet confirmed

Nvidia is reported to be pursuing a roughly $13 billion acquisition of Hugging Face, the model-sharing platform at the centre of the open AI ecosystem, but the two outlets closest to the story disagree about whether a deal has actually been signed.

The Information reported on 26 August that Nvidia had agreed to buy Hugging Face for $12.9 billion, citing a person with knowledge of the agreement. Business Insider separately reported that the companies had discussed a transaction valuing Hugging Face at more than $13 billion, but said no agreement had been signed and the talks could still fall apart.

Neither Nvidia nor Hugging Face has announced a transaction. Until one of the companies does, the precise status is unresolved: there is reporting of an agreement, reporting of unsigned negotiations and no public term sheet, closing timetable or regulatory filing.

This would not be a routine software acquisition. Hugging Face is the default distribution and collaboration layer for much of open AI: developers use it to publish, find, test and deploy models, datasets and applications. The company said in its spring ecosystem review that the platform reached 13 million users, more than two million public models and more than 500,000 public datasets in 2025.

Buying that hub would move Nvidia further up the AI stack. It already supplies the GPUs used to train and run many of the models hosted there. Ownership of Hugging Face would add the place where those models are discovered, downloaded and evaluated — and provide a direct view of which architectures and tools are gaining traction before demand reaches a data-centre order book.

That strategic position helps explain why the reported price cannot be read as a conventional software multiple. The Information put Hugging Face’s recent annualised revenue at about $150 million. A $12.9 billion purchase would therefore value the company at roughly 86 times that figure.

It would also be Nvidia’s largest completed acquisition by a wide margin if it closes. The chipmaker paid $7 billion for networking specialist Mellanox in 2020, still its biggest completed deal.

Nvidia can afford the attempt. It reported $96.2 billion in quarterly revenue for the period ended 26 July, up 106% from a year earlier, as spending on AI infrastructure continued to accelerate.

The harder question is neutrality. Hugging Face has become useful partly because rival labs, cloud providers and chipmakers can all distribute work through the same platform. Nvidia ownership would leave developers asking whether model rankings, hosted inference, optimisation tools or hardware defaults would continue to treat AMD, Google, Amazon and other competing systems on equal terms.

The same concern runs in the other direction. Open models are strategically useful to Nvidia because they broaden the pool of developers and companies that need general-purpose AI compute. Closed-model companies including Google, Amazon, Microsoft, OpenAI and Anthropic are developing their own chips or buying more custom silicon. A stronger open-model ecosystem gives Nvidia another route to keep workloads centred on its hardware.

A transaction of the reported size would also face formal review. The US Federal Trade Commission’s 2026 Hart–Scott–Rodino threshold for reportable transactions is $133.9 million, far below the proposed price. Regulators would be likely to examine whether control of a widely used model hub could disadvantage competing hardware or cloud platforms.

For now, there is no public answer on whether Hugging Face would remain operationally independent, whether its hosting and commercial terms would change or whether the reported $12.9 billion figure is final.