Nvidia’s $12.9 Billion Bet on Hugging Face Reshapes the Open AI Frontier

Clément Delangue reached for the phone this summer. The Hugging Face CEO, long insistent on independence, contacted Nvidia chief Jensen Huang. Weeks later the two companies announced a deal. Nvidia will pay roughly $12.9 billion to buy the platform that has become the default home for open-source artificial intelligence models.

The Register first captured Delangue’s shifting tone days before the formal announcement. He told the publication the “planets aligned” for a partnership and set an ambitious target: grow the user base from 18 million to 100 million. The Register reported the comments on September 3, just as rumors hardened into confirmation.

But the transaction goes far beyond alignment. It hands Nvidia control of the single largest distribution channel for open-weight models. More than 3 million models, 500,000 datasets and 1 million applications now sit under the chipmaker’s roof. Developers will still choose any hardware they want. At least that is the pledge. Nvidia’s own blog post on the deal repeated the promise three times.

“Hugging Face will remain an open platform for the entire AI ecosystem,” Huang wrote. “Developers will choose the models they want, the frameworks they want, the clouds and inference service providers they want and the computing platforms they want. NVIDIA compute will not be required.”

The words matter. Last year Hugging Face turned down a $500 million investment from Nvidia that would have valued the startup at $7 billion. Independence carried weight then. Something changed. Delangue told CNBC that he was ready for his next chapter. He and co-founders Julien Chaumond and Thomas Wolf will join Nvidia after the deal closes.

Observers immediately questioned whether true neutrality can survive. Ars Technica laid out the stakes hours after the announcement. The publication noted that Hugging Face had rejected Nvidia’s earlier overture to stay independent. Now the “GitHub of AI” answers to the company whose chips power the vast majority of serious training and inference workloads. Ars Technica detailed the shift.

Numbers tell part of the story. Hugging Face crossed $100 million in annual run rate in June, according to a post from Delangue on X. By late summer the figure reached roughly $150 million. The $12.9 billion price tag implies a multiple above 85 times current revenue. Steep. Yet Nvidia structured the deal with $11.9 billion going to shareholders and up to $1 billion in retention equity for employees. Regulators must still approve the transaction. Closing is expected in the first half of 2027.

SiliconANGLE captured fresh details the same day. The outlet reported that AMD, Intel and Qualcomm participated in Hugging Face’s most recent funding round. So did Salesforce, which itself had explored buying the company. Those investors now cash out at a handsome premium. SiliconANGLE broke down the investor implications.

The timing feels deliberate. Open-weight models have gained serious traction. Meta’s Llama series, Mistral’s releases and a wave of fine-tuned variants now rival closed systems on many benchmarks. Enterprises want options. They want to run models on their own infrastructure or in multiple clouds. Hugging Face made that easy. Its Transformers library sees millions of daily downloads. Its Spaces feature lets anyone test applications in a browser.

Nvidia already contributed heavily to the platform. The company has published more than 500 models and 250 open datasets on Hugging Face. Huang highlighted those contributions in his announcement. The acquisition, he argued, will improve reliability, safety evaluations, inference speed and deployment tools. Extra resources from Nvidia’s balance sheet should accelerate those efforts.

But. The risk of favoritism lingers. Will Nvidia-tuned models rise to the top of search results? Will competing accelerators lose visibility over time? Justin Boitano, Nvidia’s vice president of enterprise AI, tried to calm those fears. “We will have to get through all the regulatory review,” he said in a joint interview. “But we think overwhelmingly they’re going to see this as really a positive outcome.”

Observer examined Delangue’s change of heart in depth. The publication reported that the CEO initiated contact over the summer. Huang described the talks as moving quickly once Delangue signaled openness. The deal values Hugging Face nearly three times its 2023 mark of $4.5 billion. Observer explored the motivations.

Revenue context adds perspective. Most of Hugging Face’s users — 97 percent by Delangue’s earlier count — pay nothing. The company makes money from enterprise subscriptions, private repositories, inference endpoints and support contracts. It hosts hundreds of petabytes while keeping core services free. That model built trust. It also kept costs in check. The firm reportedly turned profitable in 2025.

Analysts see strategic depth. Owning the front door to open AI gives Nvidia influence over which models gain adoption. More popular models drive more compute demand. More compute demand sells more GPUs. The loop is powerful. At the same time the purchase reduces reliance on a handful of closed-model labs that have started designing their own chips.

SEC filings confirmed the structure. Nvidia disclosed the agreement on September 2. The document stresses that Hugging Face will continue to support models, datasets and applications from any source. It will keep working with other silicon vendors. Those commitments aim to ease antitrust concerns in Washington and Brussels.

Recent commentary on X reflected the split reactions. Some developers cheered the added resources. Others warned of creeping bias. One post captured the tension neatly: the open-source AI hub now has a corporate landlord. Yet the platform’s momentum remains real. From 13 million users in 2025 to 18 million today. From 2 million models to more than 3 million. Growth has not slowed.

Delangue’s 100 million user goal suddenly looks more attainable. Nvidia’s global reach, engineering talent and infrastructure could remove previous bottlenecks. The company has committed billions to startup investments and ecosystem support. This deal fits that pattern while securing a critical asset.

Questions remain. How will moderation policies evolve? Who sets priorities for new features? Can a subsidiary truly stay neutral when its parent dominates the underlying hardware? Nvidia insists the answers favor openness. History shows such promises face pressure over time.

The deal also highlights a broader consolidation wave. AI infrastructure is concentrating. Hardware leaders want software control. Software platforms want scale. Independent players find the economics difficult. Hugging Face held out longer than most. Its $150 million run rate and near-profitability gave it options. In the end the price proved too compelling.

Huang framed the acquisition as validation for open models. “The world will need both closed models and open models,” he said on a recent earnings call. “And both closed models and open models are skyrocketing in use.” The purchase puts Nvidia’s money behind that statement.

For industry insiders the implications stretch beyond one transaction. Model discovery, evaluation and deployment now flow through a single company’s platform. Enterprise adoption patterns could shift. Startup strategies may change. Developers might watch more closely where their uploads end up.

And the price. Nearly $13 billion for a company that raised less than $400 million across all prior rounds. The return for early backers is extraordinary. Sequoia, Google, Amazon and the rest of the 2023 syndicate exit at a massive markup. Even Nvidia itself, an investor in that round, buys more of what it already partly owned.

Regulators will study the competitive effects. Antitrust teams understand network effects. They know how platform control influences downstream markets. The commitment to remain open and multi-vendor will face testing. So will the practical behavior once integration begins.

Hugging Face built something rare. A trusted hub where researchers, engineers and enterprises collaborate on AI without heavy commercial pressure. That culture delivered real progress. Fine-tuned models proliferated. Benchmarks improved. Accessibility expanded. The question now is whether that spirit survives inside one of the most powerful corporations in technology.

Delangue believes it can. His team is staying. The mission continues, he posted on X after the announcement. The founders see the deal as fuel for the next phase rather than an end. Time will test that view.

For now the planets have aligned. Nvidia owns the repository that shapes how millions build with AI. The open frontier just received a very large corporate sponsor. Its independence looks different today than it did last week. The code, the models and the community remain. Who steers them has changed.