X Debate: Half an Hour After Jensen Huang Resigned, Hugging Face CEO Clem Delangue Called It "a Really Sad Day" but Vowed to "Keep Fighting for Open-Source AI, Harder Than Ever," as Sebastian Raschka and Nathan Lambert Voiced Worries About NVIDIA Nemotron and the Pending Hugging Face Deal

Background and Industry Pain Points

On the evening of November 12, NVIDIA announced that Jensen Huang had resigned as CEO, effective immediately. The open-source AI community quickly focused on two things: NVIDIA's $12.93B acquisition of Hugging Face, announced in September, has not closed yet; and NVIDIA's own Nemotron line is one of the few model efforts that publishes weights, training data and training recipes together. About half an hour after the release, Hugging Face CEO Clem Delangue called it "a really sad day" and said the team would "keep fighting for open-source AI, harder than ever." The post names no one, and replies widely read it as a response to the transition.

Architecture Highlights and Mechanism

The debate is about governance, not technology. Per NVIDIA's 8-K filing, NVIDIA committed in the deal to keep Hugging Face open, keep allowing uploads and downloads of any models and datasets, and keep supporting other silicon vendors. Those commitments sit in the deal terms, not with one CEO. But the deal only closes in the first half of 2027 after regulatory review, and the community worries priorities could shift during a leadership change. Raschka's point: open-weight models have alternatives, but Nemotron's open data and full training recipes are hard to replace.

Benchmarks and Measured Results

This is an industry discussion with no benchmarks. Verifiable numbers: Delangue's post has about 164K views and 4,800 likes. NVIDIA says it has released more than 500 models and 250 open datasets on Hugging Face, and the platform serves 18 million developers and 3 million models (vendor figures from the September deal announcement). Neither NVIDIA nor Hugging Face had posted any update on the deal at writing time.

Developer Guide

In the near term, Hugging Face Hub uploads, downloads and inference keep working, and existing Nemotron weights and datasets remain available. Teams that depend on Nemotron recipes or data can archive the exact model and dataset versions and licenses they use, and keep comparable open models on a shortlist. Worth watching next: whether NVIDIA's November 17 (US time) earnings call mentions the Hugging Face deal, and how regulatory review progresses.