Nvidia agrees to buy AI model hub Hugging Face for $12.9 billion, reports say

Nvidia has agreed to buy Hugging Face, the company whose platform hosts the internet's largest library of shared artificial-intelligence models, in a deal reported to be worth roughly $12.9 billion, according to CNBC and The Information, which first reported the terms. Neither company has issued a public statement confirming a signed contract, and people briefed on the talks cautioned that the transaction could still collapse before it is finalized. If completed, it would be the largest acquisition in Nvidia's history and would place the chipmaker in direct control of the repository that much of the AI industry uses to store, share and download open-source models.

The numbers

The reported $12.9 billion price tag is a striking reversal from where talks between the two companies began. Nvidia had previously proposed investing $500 million in Hugging Face at a valuation of about $7 billion, an offer the startup turned down, according to reporting cited by TechCrunch. The new figure values the company at more than 80 times its estimated annual revenue, a multiple typical of the AI infrastructure boom rather than of conventional software businesses. The deal talks became public the same week Nvidia reported second-quarter results showing revenue of $96.2 billion for the period ended July 26, up 106 percent from a year earlier, with data-center revenue alone reaching $89.0 billion, according to the company's earnings release and the underlying filing posted to the SEC's EDGAR system. Chief executive Jensen Huang told analysts on the earnings call, hosted through the company's investor relations site, that demand for AI computing was accelerating rather than plateauing, a backdrop that helps explain why Nvidia is willing to spend heavily to secure a foothold in model distribution as well as chips.

How Hugging Face became a chokepoint worth buying

Hugging Face began as a modest New York startup and grew into what developers commonly call the GitHub of machine learning: a hosting platform where research labs, universities and companies upload trained models, datasets and code so that anyone can download, fine-tune or deploy them. That position has made it a piece of infrastructure that much of the AI industry depends on without owning any of the underlying computing power itself, relying instead on cloud providers and chipmakers to run the models it hosts. Nvidia's interest is consistent with a broader pattern of the last two years, in which the chipmaker has taken minority stakes or forged deep partnerships with companies across the AI stack, from model developers to cloud infrastructure firms, rather than remaining strictly a supplier of graphics processing units. Owning the most widely used distribution point for open models would extend that reach from hardware into the software layer developers touch daily.

The talks also arrive amid a wider debate in Washington over whether open-weight AI models, the kind Hugging Face specializes in distributing, should face tighter export or security controls. Delangue and Nvidia's Huang were both among more than twenty technology executives who signed a letter earlier this year urging federal officials to support rather than restrict open models, arguing that openness helps American technology compete globally against alternatives developed in China. Delangue has separately said in television appearances that Hugging Face relied on an Nvidia-modified version of an open-source model to help defend its own systems after a cyberattack, an example he has cited as evidence that openness strengthens rather than weakens security. That shared advocacy for open models may help explain why a deal between the two firms was plausible in the first place, even though a sale would put a supposedly neutral hosting platform under the ownership of one of the industry's largest hardware vendors.

Who is affected, and what people are saying

The companies and developers most directly affected are the millions of engineers and researchers who rely on Hugging Face's hub to obtain models that run on hardware made by Nvidia's rivals, including AMD, as well as custom chips built by Google, Amazon and OpenAI. Those firms have a direct commercial interest in the platform remaining neutral, and are expected to raise that concern if regulators open a formal review. Hugging Face's co-founder and chief executive, Clément Delangue, has been an outspoken advocate for keeping AI models open rather than proprietary, arguing publicly this month that the security case for open models is growing. Speaking to CNBC about the broader market for open-weight systems, he said:

"AI cybersecurity is going to become a huge market in the U.S. and in the world. In this market, probably open models will be kings."

That stance sits somewhat uneasily with a sale to the world's most valuable chip company, and analysts covering the deal have flagged the tension between Hugging Face's stated commitment to neutrality and its prospective new owner's obvious interest in favoring its own hardware ecosystem. Nvidia and Hugging Face representatives did not respond to requests for comment carried in multiple outlets' reporting on the talks, and Bloomberg, which also covered the discussions, likewise noted that sources described the negotiations as still in progress rather than finalized.

What happens next

Because the deal has not been formally announced, no regulatory filings have been made public, but a transaction of this size would almost certainly draw scrutiny from antitrust authorities in both the United States and the European Union. Reviewers would likely examine whether Nvidia's ownership of a dominant model-hosting platform could be used to disadvantage rival chipmakers or cloud providers whose customers rely on Hugging Face to distribute software, an inquiry that industry analysts expect could stretch into 2027 given the pace of similar reviews of other large technology mergers. The outcome could range from unconditional clearance to commitments on interoperability and access, or, in a more contested scenario, a formal challenge to the deal. For now, the acquisition remains unconfirmed by either party, and the terms, timeline and structure could still change before any agreement is signed. Investors and developers alike are watching for an official announcement from Nvidia, whose next scheduled disclosures to the SEC would be the venue where a completed transaction of this size would eventually have to be detailed in full. An independently reported account of the still-unsigned negotiations, published the same day, likewise cautioned that terms could shift before any deal is finalized.