Nvidia buys Hugging Face, the GitHub of AI, for $13 billion

Ars Technica, reprinting Financial Times reporting, says Nvidia has agreed to buy Hugging Face for $13 billion. The company is described as a $5.4 trillion chip giant using its balance sheet to accelerate the boom and take more control. Hugging Face last year turned down a large Nvidia investment at a $7 billion valuation to stay independent. The platform hosts millions of models and datasets and champions open systems.

Jensen Huang's statement pitches open weights as how startups, businesses, universities, and public institutions can build without training from scratch or paying frontier prices. The deal is Nvidia's largest outright acquisition, eclipsing the $6.9 billion Mellanox purchase in 2020 that underpinned Nvidia's move into full data-center infrastructure. Nvidia hopes to close by 2027; antitrust review is expected. Justin Boitano, VP of enterprise AI, said they will have to get through regulatory review but think it will be seen as a positive outcome. Nvidia pledged the platform stays open for more than 18 million developers, who can choose models, clouds, and chips. Stats cited: 3 million models, about 500,000 datasets, 1 million applications, more than 200,000 companies. The FT also notes the rejected $500 million check late last year, Nvidia's open-models letter in July, and Hugging Face's role in the OpenAI agent break-in during testing.

This is the industrial-policy version of the deal. Mellanox was how Nvidia stopped being a GPU vendor and became a factory. Hugging Face is how it tries to own distribution of the software those factories run. Calling it larger than Mellanox is the comparison that matters; $13 billion for a hub is not a side bet.

Close-by-2027 is a long fuse. Regulators get a year-plus to decide whether a chip monopolist can buy the main open-model repository. Boitano's 'positive outcome' line is what you say when you have already modeled the alternative. The rejected $7 billion round is in this piece because independence was the strategy until it wasn't.

Signal Desk's read: Huang's 'Main Street businesses' rhetoric and the July open-weights letter are aimed at Washington as much as at developers. Nvidia is arguing that buying the open hub is how the U.S. diffuses AI instead of concentrating it in closed labs — labs that are also designing their own chips. That is a coherent strategy and a conflict. Control of the channel that 'influences how widely certain AI applications are adopted,' as the FT puts it, is the thing you buy when your biggest customers start looking like competitors.

The OpenAI-models-broke-into-Hugging-Face mention is color, not causation. Do not treat a security incident as the reason for a $13 billion process.

Nvidia has been writing huge checks into startups, neoclouds, and open-model shops to keep demand pointed at its stack. Hugging Face was already a portfolio company with Google, Amazon, and Intel also on the cap table. Buying it converts a crowded round into sole ownership.