Written by Admin Alex · Fact-Checked by M.Ali · Info Verified September 2026
We review and update this article regularly as new information becomes available.
TL;DR: Nvidia is buying Hugging Face, the open repository that hosts millions of AI models and datasets, for $12.93 billion. Hugging Face turned down a $500 million offer from the same company last year. Now it’s saying yes, and the terms say the platform stays open to everyone, not just Nvidia hardware.
Three years ago, Hugging Face said no to Nvidia’s money. On September 3, it said yes, at more than 25 times the price.
The deal values Hugging Face at $12.93 billion, a number that would have sounded absurd back in 2023 when the company last raised outside capital, a $235 million round led by Salesforce Ventures. Total funding before this deal: $395 million. Annualized revenue: roughly $150 million. Do the math and Nvidia just paid something like 86 times revenue for a company that mostly gives its core product away for free.
That’s because the product was never really the point. Hugging Face is infrastructure. It hosts about 3 million AI models, a million applications, and half a million datasets, used by more than 18 million developers who treat it the way web developers treat GitHub. If you’ve fine-tuned an open-weight model in the last four years, there’s a good chance you pulled it from there.
Jensen Huang framed the acquisition as protective rather than possessive. “Hugging Face will remain an open platform for the entire AI ecosystem,” he said, adding that “Nvidia compute will not be required to build on or deploy through Hugging Face.” Whether that promise survives five years of shareholder pressure is a different question, but for now it’s the official line.
Hugging Face’s own CEO, Clem Delangue, gave a more practical explanation for why he finally said yes. “It needs more compute, more support, more collaboration, and more visibility,” he said. “That’s why we went to talk to Jensen.” Running the internet’s default AI model hub is apparently an expensive habit, and Nvidia has the deepest pockets in the industry to fund it.
Why Nvidia wants this
Two motives stand out. First, bundling. Nvidia sells enormous amounts of GPU capacity to enterprise customers, and pairing that hardware with the platform where those customers already go to find and deploy models is an obvious sales lever. Second, open-weight strategy. Nvidia has contributed more than 500 models to Hugging Face already, and owning the distribution channel for open models gives it a say in how the open-source side of AI develops, at a moment when Chinese labs like DeepSeek and Moonshot are shipping competitive open models of their own.
There’s also a simpler read: control of the plumbing. Whoever owns the place where the AI industry stores and shares its models has a permanent seat at the table, regardless of which specific lab wins any given benchmark this quarter.
What could go wrong
Hugging Face’s value came from being neutral ground, a place OpenAI, Anthropic, Google, Meta, and thousands of independent researchers all used without worrying whose team it was on. That neutrality is now owned by the world’s largest AI chipmaker, a company with obvious commercial interest in which hardware people use to run the models it hosts. Huang’s promise not to require Nvidia compute is worth watching over the next year, not just taking at face value.
Regulators haven’t weighed in publicly yet, but a deal of this size, from a company already facing multiple antitrust inquiries elsewhere, tends to attract scrutiny eventually.
Bottom Line
Nvidia didn’t just buy a company. It bought the default home page of the open-source AI world, for a price that only makes sense if you think owning that home page is worth more than the business itself. Given how the last few years of AI infrastructure land grabs have gone, that’s not a crazy bet. It’s just a big one.
