SANTA CLARA, Calif.—The world’s dominant seller of artificial-intelligence chips is in talks to buy the site where many of those chips get their software.
Nvidia is in advanced discussions to acquire Hugging Face, the leading platform for open-source AI models, and a deal could be reached as soon as this week, according to people familiar with the matter. Bloomberg reported the talks. Terms have not been finalized and the timing and shape of any agreement could still change, the people said. Representatives of both companies declined to comment.
Under the terms being discussed, Nvidia would pay $12.9 billion for Hugging Face plus roughly $1 billion in employee-retention incentives, for a total of about $14 billion, the people said. The purchase price amounts to more than 80 times Hugging Face’s annual revenue of about $150 million, a multiple that reflects the premium buyers pay for the distribution layer of AI rather than for the revenue it currently generates.
If completed, the acquisition would be the largest in Nvidia’s history, surpassing the $6.9 billion it paid for the networking company Mellanox in 2020. It would also rank among the biggest deals yet in the AI boom, and it would put a distinctly open-source institution under the roof of a chip maker whose fortunes rest on proprietary hardware.
Hugging Face was founded in 2016 and has become the default doorway to open AI. Developers use its hub to download models from Meta, Google, and thousands of independent researchers; its libraries are woven into the way much of the industry builds AI applications. The company also sells enterprise services, hosting models and running inference for paying customers, a business whose growth has attracted the interest of rivals far larger than itself.
The company has been courted before and said no. In late 2025, Hugging Face rejected a $500 million investment that valued it at about $7 billion, according to people familiar with the matter, choosing to keep its independence rather than take a strategic investor’s money. That the founders are now negotiating a sale at roughly double that valuation suggests the conversations with Nvidia reached a point earlier approaches did not.
The strategic logic is straightforward. Nvidia sells the accelerators on which most AI models are trained and run, and it wants to control more of the path from model to deployment. Hugging Face sits at the start of that path: it is where developers discover, test, and download open-weight models, and increasingly where enterprises pay to serve them. Owning the doorway would let Nvidia steer the open-source ecosystem toward its own hardware, its own software stack, and its own cloud services rather than leaving that traffic to run on a competitor’s systems.
The deal would also answer a quiet threat. Hugging Face has built its inference business to run across clouds from multiple providers, and its platform is where models that could one day compete with proprietary systems are distributed. A neutral hub is valuable precisely because it is neutral; under Nvidia, rivals may wonder whether the doorway now leans toward one vendor’s stack. Nvidia will need the company’s community to keep trusting the platform for the acquisition to pay off.
The roughly $1 billion retention package acknowledges that risk. Hugging Face’s value sits in a small team of engineers and in the goodwill of a developer community that has prized the company’s independence. Chief executive Clement Delangue, who co-founded the company, has built a brand around openness, and whether he stays and how the company is integrated will shape whether developers treat the sale as a natural next step or a betrayal.
Regulators may have something to say first. Nvidia’s dominance of AI chips has already drawn scrutiny in the United States and Europe, and an acquisition that puts the leading open-model hub under the same roof could invite questions about whether the company is extending its control from silicon into software. Nvidia knows that terrain well: its proposed $40 billion takeover of the chip designer Arm collapsed in 2022 after regulators objected.
Financially, the purchase is well within reach. Nvidia generates cash at a scale that makes a $14 billion acquisition a rounding error against its annual results, and investors have shown little appetite for questioning its strategic bets while its core business keeps growing. The richer question is whether the price makes sense for a company earning about $150 million a year.
The answer depends on where the AI industry is headed. If open-weight models continue to spread, the hub where they live becomes more valuable with every download, and the enterprise business Hugging Face is building could grow into the revenue the multiple assumes. If the field consolidates around a few proprietary systems, the platform’s pull could fade.
For the developer who wakes up tomorrow and downloads a model, nothing changes at first. The site looks the same, the model cards read the same, and the code works the same way. Only the owner is different—and, if the deal closes this week, the doorway to open AI will belong to the company that makes the chips most of it runs on.
