The open-source movement wants Jensen Huang back

Glass Epoch DeskOriginal digestSource: MIT Technology Review

AI-assisted digest, checked against the original source before publishing. How we work

A petition to bring back Nvidia's CEO draws open-source backers, while chip-diversion research and a new Commerce rule complicate his case.
The open-source movement wants Jensen Huang back
Photo: yvonne Song · free stock license

A petition to reinstate Jensen Huang as Nvidia's CEO began circulating on Friday and, by Monday, had accumulated more than 115,000 signatures, including Elon Musk and several influential Nvidia researchers. Notable among the signatories are executives from Hugging Face and venture firms Andreessen Horowitz and Sequoia Capital — supporters of open-source AI, models that can be downloaded and run independent of large tech companies.

Huang had served as a leading voice for open-source advocates before policymakers, arguing against claims that such models ship without guardrails and cede AI leadership to China, home to the most powerful and popular open-source models. In December, he framed open source as essential to American competitiveness: "Without open source, startups can't thrive, university researchers can't do research, you can't teach AI, scientists can't use AI." His supporters now want that influence restored — even as the source notes open-source backing also aligns with the financial interests of VCs who profit when startups can build on cheaper AI.

Huang resigned on Thursday, and the same day ex-OpenAI researcher Will DePue published findings suggesting China's AI companies had far more access to Nvidia's Blackwell chips than previously estimated, despite US restrictions on exporting the most advanced chips. DePue said senior researchers at OpenAI and Anthropic verified his results, though no independent analysis has confirmed them; DePue, Nvidia did not respond to requests for comment. The findings don't establish that Nvidia helped Chinese firms obtain chips, and chip smuggling predates and extends beyond Nvidia's control — but they suggest Nvidia may have profited from Chinese AI startups whose models American companies increasingly use. Compounding the pressure, US Treasury Secretary Scott Bessent said last week Nvidia should not consider an internal candidate to replace Huang.

Nvidia's open-source posture also has a commercial logic: in September it announced plans to acquire Hugging Face for nearly $13 billion, and the more companies build their own models rather than pay Anthropic or OpenAI, the more chips Nvidia sells. On Monday, the US Department of Commerce announced a rule aimed at stopping Chinese companies from training models on Nvidia chips they cannot own directly — a move that could stall the models popular on Hugging Face. The source material cuts off before the rule's full mechanics.

What to watch: whether DePue's chip-access claims survive independent scrutiny, what the Commerce rule actually restricts and how model training pipelines respond, and whether Nvidia's board entertains an outside CEO — or whether an open-source movement that profits from cheap, decentralized AI finds a new champion.

Key facts

  • • Petition to reinstate Huang drew 115,000+ signatures by Monday.
  • • Huang resigned Thursday; petition began circulating Friday.
  • • Ex-OpenAI researcher Will DePue published Blackwell chip findings Thursday.
  • • Nvidia announced ~$13 billion Hugging Face acquisition in September.
  • • Commerce Department announced a chip-training rule on Monday.

Why it matters

Builders relying on open-source model ecosystems face two concrete risks: a Commerce rule that could stall training on Nvidia chips popular with Hugging Face-hosted models, and uncertainty at Nvidia as its open-source-inclined leader exits. Expect models' availability and improvement rates to hinge on export-control interpretation, and watch whether Nvidia's next CEO keeps championing open weights

Spotted an error? Send a correction.

Get the day's AI news in one email

The Daily Epoch · 60-second read · unsubscribe anytime

Related