Hugging Face’s Delangue vows to keep fighting for open source as Nvidia’s deal loses its champion

In September Nvidia agreed to buy Hugging Face, the main hub for open AI models, for $12.9 billion (deal coverage). TechCrunch reports the deal is still slated to close in the first half of 2027. It would be Nvidia’s second-largest acquisition, behind its $20 billion purchase of Groq’s assets in December 2025. The executive who championed it, and who pushed hard for open-weight AI in Washington, is now gone.

Hugging Face CEO Clément Delangue called it a really sad day and promised to keep fighting for open source harder than ever. The subtext: people who rely on open models are asking whether Nvidia’s commitment was company strategy or one leader’s conviction (analysis). TechCrunch’s requests for comment on whether Hugging Face had spoken with interim CEO Colette Kress went unanswered.

Open-weight models are among the most useful things in AI right now. They let researchers inspect systems, let companies run models on their own infrastructure and keep the market from collapsing into a few closed providers. Nvidia’s former CEO made that case in a July policy letter signed by 25 companies, arguing open models strengthen competition, safety and sovereignty. Losing that voice at the top of the most important AI hardware company is a real setback for the open camp, one of the five losses TechCrunch lists.

TechCrunch notes the Justice Department has been asking how the acquisition would affect competition, because it gives Nvidia control of a major distribution channel for AI models. With a strong founder in charge, the bet was that his values would protect Hugging Face’s neutrality. Without him, the deal looks more like a chip giant owning the shelf where everyone else’s models sit (earlier analysis). That concern deserves sharper scrutiny, not less.

There are reasons for some optimism. Nvidia’s open work runs deeper than one person: TechCrunch points to the $6 billion Poolside deal to build an American open AI ecosystem and the expanded Nemotron open-weight family, including a 120-billion-parameter model for enterprise agents. Selling chips is easier when many people build and run many models, so openness still serves Nvidia’s interests.

The two tests: whether interim leadership publicly reaffirms the open-model strategy, and whether Hugging Face gets binding neutrality commitments before closing. A cheerful post on X is not governance. If the platform’s openness rests only on goodwill, the open-source community should push for more. More in our Nvidia coverage and startups coverage.