KEY POINTS
- A petition to reinstate Jensen Huang as Nvidia CEO had more than 115,000 signatures by Monday, including many open-source AI backers.
- Unconfirmed research suggests Chinese AI firms had far greater access to Nvidia Blackwell chips than estimated.
- A new US Commerce rule targets remote access to Nvidia chips used by Chinese companies to train models.
A petition calling for Jensen Huang to be reinstated as Nvidia CEO began circulating on social media on Friday and had collected more than 115,000 names by Monday, MIT Technology Review reported. Huang resigned on Thursday. Signatories include Elon Musk and several influential Nvidia researchers.
According to the outlet, a notable share of the signers come from the open-source AI community. They include senior figures at Hugging Face, a platform for sharing AI models, and at the venture capital firms Andreessen Horowitz and Sequoia Capital. These people have a stake in models that can be downloaded and run independently of large tech companies.
MIT Technology Review describes Huang as a leading voice for open source before policymakers. He argued against worries that open models lack guardrails and hand AI leadership to China, where many of the most powerful and popular open models are built. Supporters now want that advocate back in his old role.
The effort faces new obstacles. On Thursday, the day Huang stepped down, former OpenAI researcher Will DePue published findings suggesting that Chinese AI companies had much greater access to Nvidia's advanced Blackwell chips than previously thought. China is barred from obtaining those chips as part of efforts to protect the US lead in AI.
How the companies might have obtained the hardware is unclear. DePue says senior researchers at OpenAI and Anthropic verified his results, but no independent analysis has confirmed them. He did not respond to the publication's interview request, and Nvidia did not immediately comment. The report stresses that the findings do not show Nvidia helped Chinese firms get the chips, and that chip smuggling was already known to reach beyond the company's control.
Even so, the research puts Huang in an awkward spot. Chinese open-source models are increasingly used by American companies, and the findings hint that Nvidia may have profited from the rise of the startups behind them. The report adds that US Treasury Secretary Scott Bessent said last week Nvidia should not pick an internal candidate to succeed Huang, which already made a return look unlikely.
A second threat comes from policy. The US Department of Commerce announced a rule on Monday meant to stop Chinese companies from training models on Nvidia chips that they cannot own directly but can reach remotely through data centers in countries such as Malaysia. MIT Technology Review says that workaround has helped Chinese open-source models improve so quickly, so the rule could slow the progress of models popular on Hugging Face.
Nvidia announced in September that it would acquire Hugging Face for nearly $13 billion. The outlet notes that the company's support for open source also has a clear commercial logic: the more businesses build their own models rather than paying OpenAI or Anthropic, the more chips Nvidia can sell.