Hajra Naz
Nvidia has agreed to acquire Hugging Face, the platform where most of the world’s open AI models are shared, for $12,930,300,000. Chief executive Jensen Huang announced the deal on 3 September 2026 and committed to keeping the platform open, writing that “NVIDIA compute will not be required to build on or deploy through Hugging Face.” Nvidia already sells the accelerators that train and run those models, so the purchase gives one company both the hardware layer and the main place developers go to find and share what runs on it.
The deal had been reported as close for a week. In late August, beingguru covered the reports that Nvidia was moving toward a Hugging Face acquisition. What changed on Thursday is that the price and the commitments are now on the record.
What $12.93 billion buys
Nvidia’s own numbers describe the scale of what it is taking on. More than 18 million developers, researchers and creators use Hugging Face. They share more than 3 million models, 500,000 datasets and 1 million applications. More than 200,000 companies use the platform to discover, evaluate, customise and deploy AI.
For context on what that price means to the buyer, Nvidia reported revenue of $96.2 billion for the quarter ended 26 July 2026, up 106% from a year earlier. The Hugging Face purchase is worth roughly one seventh of a single quarter’s revenue.
The commitments Nvidia put in writing
Huang’s announcement sets out four specific promises. Hugging Face will remain an open platform for the entire AI ecosystem. Developers will choose their own models, frameworks, clouds, inference providers and computing platforms. The platform will keep supporting open source and open weight models from every model builder, not only Nvidia’s. And it will keep supporting multi-cloud and multi-accelerator development, so work built there can run on hardware Nvidia did not make.
Nvidia also states that it is the largest contributor of open models and data to Hugging Face, having released more than 500 models and more than 250 open datasets on the platform. That record is the main argument the company is offering that it will not close the door behind it.
Why an American chip deal shows up in your workflow
If you build with AI on a tight budget, and most independent operators in Pakistan and the Gulf do, open weight models are the reason the budget works at all. You download the weights instead of renting an API. You run them on a machine you already own or on a GPU rented by the hour. You fine tune on a client’s data without shipping that data to a vendor. There is no per token bill growing quietly in the background while you sleep.
Hugging Face is the shelf all of that sits on. It is where you find a model small enough to run locally, read what licence it carries, check whether anyone has quantised it to fit your hardware, and download the dataset to test it against. Losing that, or watching it narrow, would not be an abstract industry story. It would raise the floor price of doing AI work independently.
Nvidia’s stated position is that none of this changes. The commitments above are unusually specific for an acquisition announcement, which suggests the company understood exactly which worry it needed to answer. Hugging Face chief executive Clement Delangue told CNBC he approached Huang first and framed the deal as a way to grow the community rather than fence it in, saying the aim is to reach 100 million builders on the platform.
The part that cannot be checked yet
Openness here is a promise from a press announcement, not a term any user can inspect. Three things will show whether it holds, and all three are visible from the outside. Whether models tuned for non-Nvidia hardware keep equal billing in search and rankings. Whether third party inference providers stay listed alongside Nvidia’s own. And whether the free tier that lets a student in Multan download a model at no cost stays free once the integration work begins.
None of that will be clear for months. Until then the sensible move for anyone whose work depends on open models is the same one that has always applied: keep local copies of the weights and datasets you actually rely on, and do not assume any platform will host them forever.
Sources: NVIDIA, “NVIDIA to Acquire Hugging Face” (company announcement by Jensen Huang); NVIDIA Q2 fiscal 2027 results; CNBC.
