Nvidia CEO Jensen Huang has come out against US efforts to block Chinese open-source AI models, arguing that restricting them could actually weaken American security. He says models like Kimi, developed by Chinese startup Moonshot AI, are performing well and that US companies "should definitely use them."
Huang suggests that enterprises can download these models and run them in secure sandboxes, maintaining full control over their data and access. Open weights also allow external researchers to inspect for vulnerabilities. Relying on a single dominant model, he argues, creates a single point of failure.
He also dismisses fears that Kimi will crush OpenAI or Anthropic. Low-cost and even free AI models expand the market, driving demand for more chips, data centers, and computing power. "Free AI is good for hardware," Huang says bluntly.