Nvidia has agreed to acquire Hugging Face, the largest open platform for building and hosting machine learning models, in a deal worth roughly $12.9 billion. The Hugging Face acquisition is Nvidia’s most direct financial commitment yet to open-weight AI, a category the company has spent years championing in public while selling the hardware that trains and runs the models underneath.
The total splits into two parts: $11.9 billion goes to Hugging Face’s existing investors, with a separate equity-based retention pool of up to $1 billion set aside for employees who join Nvidia, according to Reuters figures reported by PC Gamer. That structure suggests Nvidia is paying almost as much to keep Hugging Face’s engineering and community team in place as it is for the platform itself.
Why Nvidia says the platform stays open
The obvious worry is that owning Hugging Face gives Nvidia leverage to steer developers toward its own hardware and software stack. Nvidia CEO Jensen Huang addressed that directly in the company’s announcement: “Hugging Face will remain an open platform for the entire AI ecosystem. 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 to build on or deploy through Hugging Face.”
Huang also committed to continued support for both open source and open weight models on the platform, the two categories that have made Hugging Face the default hosting point for everything from research checkpoints to production language models. For a company that also sells the GPUs behind closed, cloud-hosted alternatives, keeping that neutrality credible matters more than the purchase price.
Announced or already done?
The coverage itself is not consistent on how far along this actually is. PC Gamer describes Nvidia as intending to acquire Hugging Face; GamesIndustry.biz states flatly that Nvidia “has acquired” it. Nvidia’s own announcement, however, is titled “Nvidia to acquire Hugging Face”, future tense, which points to a signed agreement rather than a closed deal. Acquisitions at this size typically still need regulatory clearance before they finish, so it is worth treating this as agreed rather than completed until Nvidia says the deal has closed.
What to watch next
Two things are worth tracking from here: whether Nvidia confirms the deal has actually closed, and whether Huang’s promise holds up once Hugging Face is inside Nvidia, specifically whether non-Nvidia hardware paths and inference providers keep getting equal billing on the platform rather than being quietly deprioritised.
Image: Coolcaesar via Wikimedia Commons, licensed under CC BY-SA 4.0.
