Jensen Huang urges no ban on Chinese AI as Kimi K3 triggers 12% selloff

Nvidia's Jensen Huang said the US should not restrict Chinese open-source AI models, arguing a ban would hurt American companies more than China, as chip stocks fell 12.5% in a week after Moonshot's Kimi K3.

"These Chinese models are excellent. Open-source models that are excellent should be used," Huang said in an interview with Axios on July 22 at the opening of a Wistron plant in Fort Worth, Texas, that builds Nvidia's AI infrastructure.

Moonshot AI released Kimi K3 on July 16, a model with 2.8 trillion parameters and a 1-million-token context window. Independent evaluators ranked it third on Artificial Analysis' Intelligence Index, behind only Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol, and first on LMArena's Frontend Code Arena. The model costs $15 per 1 million output tokens, compared with $50 for Anthropic's Fable 5 and $0.87 for DeepSeek V4.

"The market misunderstood the impact of DeepSeek the first time, and it has misunderstood the impact of Kimi again this time," Huang said, arguing that cheaper, more capable models drive broader adoption and, in turn, greater demand for Nvidia's chips and data center infrastructure. Nvidia briefly lost its crown as the world's most valuable company after the Kimi K3 release, with semiconductor stocks dropping more than 20% from their June 2026 peaks.

Washington Weighs Sanctions as IP Theft Allegations Mount

White House Office of Science and Technology Policy Director Michael Kratsios accused Moonshot AI on July 22 of using Anthropic's Fable model to develop Kimi K3, claiming the company built a "sophisticated internal platform" to perform large-scale distillation while evading detection. Treasury Secretary Scott Bessent said the administration is investigating whether Chinese AI companies improperly extracted knowledge from US models, with sanctions possible within "days or weeks."

Huang dismissed the theft narrative. "Distillation, learning from AI, learning from other sources of knowledge, is fundamental to intelligence," he said. He also argued that open-weight models are safer than closed ones because researchers can inspect them for vulnerabilities, adding that "if everything just becomes one single model, one single point of attack, one single source of failure, I think the world is much, much more vulnerable."

Silicon Valley Rallies Against a Ban

Nearly 200 Silicon Valley companies, including startup accelerator Y Combinator and encrypted email provider Proton, urged the Trump administration not to restrict access to Chinese open-weight AI models. In letters sent by the newly formed Little Tech Association, the companies argued that US startups rely on affordable open AI models that cannot easily be replaced by more expensive proprietary alternatives from OpenAI or Anthropic.

Tesla Chief Executive Officer Elon Musk backed Huang's stance, replying "True" to a social media post highlighting the Nvidia leader's argument. Huang also called on Anthropic to open up its cybersecurity model Claude Mythos, which is currently restricted to a vetted group of roughly 200 partners through its Project Glasswing program.

The full weights of Kimi K3 are set to go public on July 27, at which point independent testing will either confirm or challenge Moonshot's benchmark claims. Epoch AI estimates the gap between the best open and closed models has narrowed to about three months, raising the question of whether US restrictions would have any practical effect once the weights are available globally.

Nvidia shares, which have declined along with the broader semiconductor sector, face an uncertain regulatory outlook. Huang has endorsed keeping Nvidia's most advanced Blackwell and Rubin architectures out of China, but his broader argument — that open AI drives chip demand — will be tested as Washington decides whether to impose restrictions on Chinese models.

This article is for informational purposes only and does not constitute investment advice.