Nvidia is buying Hugging Face for about $12.9 billion, a deal that puts the chipmaker in control of the single most important distribution hub for open AI models. Jensen Huang announced the acquisition on September 3, 2026, with roughly $11.9 billion in purchase price plus up to $1 billion in stock to retain staff. The deal won’t close until the first half of 2027 and still needs regulatory sign-off.
The platform’s scale is hard to overstate: more than 18 million developers, 3 million models, 500,000 datasets, and 1 million applications, with over 200,000 companies using the service. Huang insists nothing changes — Hugging Face stays open and hardware-neutral, no Nvidia chips required. That promise is probably sincere in the short term. But the strategic logic here is blunt: the big AI labs are all building their own silicon now. Google, Amazon, OpenAI, and Anthropic have each moved to reduce dependence on Nvidia. Open-model users, by contrast, run workloads across clouds, universities, and government agencies — and they don’t fabricate their own accelerators. Huang summed it up for Axios with one line: “Free AI should be great for hardware.”
Hugging Face isn’t just a model repository either. It rents out compute time to run those models, which makes it a direct sales channel for Nvidia’s cloud partners. Nvidia scaled back its own DGX Cloud precisely because it didn’t want to compete with its biggest buyers. Now it gets a developer hub that routes jobs to partners like CoreWeave, Lambda, and Nebius through its Lepton marketplace. With $36 billion in AI cloud partner commitments reported in late July, a built-in demand engine starts to look very useful.
There’s an irony worth sitting with. Nvidia is already the largest contributor of open models on Hugging Face, with over 500 models and 250 datasets, and it runs the Nemotron Coalition alongside Mistral AI, Perplexity, and others. Now it owns the showcase where that competition plays out. Whether a hardware vendor can truly run a neutral platform for years — especially when open models from Chinese labs like Kimi K2.6 and GLM-5.1 keep outperforming Western alternatives — is the question regulators and developers will both be asking between now and 2027.
