Nvidia CEO says ban Chinese open-source AI models — embrace them instead

Jensen Huang, Nvidia's CEO, challenged Washington's emerging consensus on Tuesday in an interview with Axios, arguing that Chinese open-source AI models should not be banned but rather welcomed by the U.S. His remarks come amid a growing panic over capabilities like those demonstrated by Moonshot AI's Kimi K3, released July 16, which has proven competitive with Anthropic's most advanced publicly available model, Fable 5, in coding tests.

The White House has reportedly considered using executive power to restrict U.S. companies' adoption of Chinese AI models, including DeepSeek, Alibaba, and Moonshot AI's offerings. The feared mechanisms include government surveillance via the models themselves and economic harm to American AI companies. According to reporting, the White House considered forcing U.S. companies that use Chinese models to guarantee security and accept liability in case of breach. Huang dismissed the surveillance concern as a "misconception," emphasizing that open-source models are downloadable and their guardrails are customizable — not a hidden backdoor.

The pricing disparity illustrates why the debate matters for real business decisions. Kimi K3 costs $15 per 1 million output tokens, substantially lower than Anthropic's Fable 5 at $50, though more expensive than another Chinese open-source option, DeepSeek V4, at 87 cents for the same volume. American companies, including the AI coding startup Cursor, are already turning to these low-cost alternatives to avoid the high costs of U.S. models. Huang's argument was unambiguous: "These Chinese models are excellent. Open-source models that are excellent should be used." He argued further that the world needs both open-source and closed models, and that restricting either would be counterproductive. More AI use globally, he reasoned, will drive higher demand for Nvidia's computing hardware and infrastructure, benefiting the broader industry. The White House did not immediately respond to requests for comment on Huang's position.

Huang's intervention represents a striking contrast to the escalating anxiety in Washington over Chinese AI capabilities. The White House has reportedly explored executive action to impose security conditions and liability requirements on U.S. companies using Chinese models, driven partly by surveillance concerns and partly by worry that low-cost alternatives could undercut homegrown AI businesses. Kimi K3's release last week — capable enough to compete with Anthropic's top public model — crystallized those fears.

Huang's rebuttal centers on two claims: first, that open-source models are not inherently a security risk (they are downloadable and customizable, not a covert vector), and second, that Chinese and American models can coexist productively. His logic is blunt economic self-interest: more AI use anywhere means more compute demand, and Nvidia supplies the chips that run all AI, whether open or closed. The gap in pricing — Kimi K3 at $15 per million tokens versus Fable 5 at $50 — explains why companies are switching, and Huang's position effectively acknowledges that restriction would be economically futile and strategically backward.

No comments yet. Be the first to share your thoughts!