SAN FRANCISCO — Every time a developer pulls an artificial intelligence model from Hugging Face — the open-source hub that has become as essential to AI as GitHub is to software development — they are downloading something that may soon belong to Nvidia.
The chipmaker has agreed to acquire Hugging Face for $12.9 billion, The Information reported late Wednesday, citing a source with knowledge of the matter. Neither company responded to requests for comment, and Fortune confirmed that no signed agreement had been reached at the time of publication. The terms of the reported deal are already reshaping how the industry thinks about where power sits. Nvidia controls an estimated 90 percent of the global market for AI training chips. Hugging Face controls the platform where 13 million developers share, discover, and deploy the models that run on those chips. Owning both would complete a vertical stack no single company has yet assembled.
Nvidia already sells the shovels. It now wants to own the marketplace.
Hugging Face was founded in 2016 as a chatbot company before pivoting into what became the dominant open-source AI model repository. Its library holds more than 900,000 models — from large language models to specialised vision and audio tools — and is used by research labs, startups, and enterprises that lack the resources to build comparable infrastructure from scratch. The company’s valuation has climbed steeply. A $235 million funding round in 2023 valued it at $4.5 billion, with Nvidia among the investors. By late 2025, Nvidia’s initial acquisition discussions put the price at roughly $7 billion, according to earlier reports. The $12.9 billion figure represents a near-doubling in under a year, a premium that reflects how central Hugging Face has become to the industry Nvidia profits from.
What the deal offers Nvidia is structural, not merely financial. AI models trained on Nvidia GPUs are disproportionately distributed through Hugging Face. An acquisition would complete a chain that begins at the data centre, runs through the chip, and ends at the platform where the finished model reaches developers who cannot afford their own infrastructure. Nvidia has been building leverage at every layer of the AI value chain well before this acquisition, as its move into financing arrangements with AI cloud customers already demonstrated. A Hugging Face deal would represent the furthest extension of that approach yet.
The regulatory path will not be swift. Under the Hart-Scott-Rodino Act, a transaction of this size triggers mandatory pre-merger notification with the Federal Trade Commission and the Department of Justice in the United States; both must review the filing before the deal can close. In the European Union, the deal is almost certain to attract scrutiny under Article 22, the provision that allows Brussels to scrutinise acquisitions regardless of whether the parties meet standard revenue thresholds. Nvidia knows what that process looks like. TechCrunch reported that its $700 million acquisition of AI infrastructure startup Run:ai in 2023 drew an Article 22 referral from Italian regulators, and Nvidia challenged the Commission’s authority in court while the review proceeded. The deal eventually cleared in December 2023, but the legal fight set a precedent: Nvidia is willing to litigate the process itself, not just its outcome. At $12.9 billion, with Hugging Face at the centre of global AI development, the Commission’s interest this time will be considerably sharper.
AMD and Intel, alongside the custom-chip teams at Google, Amazon, and OpenAI, are expected to register objections during any review. All of them have development relationships with Hugging Face precisely because the platform operates outside any single hardware ecosystem. If Nvidia owns it, that neutrality is no longer guaranteed. AMD’s $5 billion chip partnership with Anthropic made clear how fiercely rival chipmakers are competing for influence over the infrastructure layer of AI, and how much they have at stake if Nvidia controls where models are distributed.
Inside the AI research community, the concern is different and harder to answer. Hugging Face has positioned itself as the open-source counterweight to proprietary AI systems: a place where anyone can publish a model and anyone can download one, free of commercial pressure to favour particular hardware. Clement Delangue, the company’s chief executive, has been explicit about that mission. Neither Nvidia nor Hugging Face has said whether Delangue would remain after an acquisition, what governance structures would protect the platform’s neutrality, or how hosting policies might change. Those questions will hang over the deal through however long the regulatory process takes.
A signed agreement had not been reached at the time of reporting. The deal could still fall apart. But if it closes, the years when the infrastructure for sharing and accessing AI models existed outside any single company’s control will have ended not with a dramatic collapse, but with a $12.9 billion transaction that neither party has yet confirmed.
Related Posts
