Nvidia's Jensen Huang is telling Washington it has the threat calculus backwards. In an exclusive interview, the chipmaker CEO defended Chinese open-source AI models as beneficial to the U.S. economy and warned that efforts to restrict them could actually damage American security.
"These Chinese models are excellent," Huang said Tuesday. "Open-source models that are excellent should be used." He argued there is "zero possibility" that China could run American companies off the road, directly challenging the growing push from Washington and U.S. AI labs to block access to cheaper Chinese alternatives.
The comments arrive as financial markets remain jittery over the release of Kimi K3 by Beijing-based Moonshot AI. The model combines near-frontier performance with dramatically lower prices and openly available weights that developers can customize, triggering fresh stock sell-offs and reigniting fears that cost-effective Chinese rivals could undermine the case for massive infrastructure spending.
Huang rejected this logic entirely. He argued that cheaper, freely available AI models would expand the overall market by introducing more people and businesses to the technology. Most users, he suggested, will still pay for the convenience, reliability and superior performance of premium closed services. The net effect benefits chip manufacturers like his own company.
"Free AI should be great for hardware," Huang said. "Free AI should be great for chips. Free AI should be great for data centers."
The Nvidia chief also took aim at Treasury Secretary Scott Bessent's suggestion that Washington might impose sanctions on Chinese AI models for allegedly stealing U.S. intellectual property. Huang characterized model training as a natural part of how intelligence develops.
"Distillation, learning from AI, learning from other sources of knowledge, is fundamental to intelligence," Huang told Axios. While he acknowledged that companies should face consequences for violating privacy or contracts, his broader message was clear: target specific misconduct, not the technology itself.
On the national security question, Huang pushed back against the argument that allowing downloaded Chinese models creates a security backdoor. Companies can customize these models and operate them in controlled environments, he said, reducing the actual risk. He went further, arguing that restricting access to advanced models makes the world less secure overall because open models allow outside researchers to identify and fix vulnerabilities.
"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," Huang said. He even criticized American companies for keeping their most powerful capabilities locked down, suggesting Anthropic should make its advanced Claude Mythos model more widely available rather than rationing access. "Holding Anthropic back is not in the benefit of the United States," he said.
Huang's framing stands apart from how senior Trump administration officials have cast AI competition with China. Where Washington sees an urgent race with clear winners and losers, Huang views AI development as an indefinite process that both countries will participate in for decades.
His remarks came at a Nvidia facility in Fort Worth where Taiwan-based Wistron manufactures AI infrastructure, a visual reminder that sophisticated chip production happens in America. Yet Huang's message undercuts the rationale behind efforts to wall off U.S. technology. The competition, in his view, cannot be won through restriction.
Author James Rodriguez: "Huang is making the hard case Washington doesn't want to hear: that trying to cage Chinese AI through policy will backfire while doing nothing to stop the underlying innovation that scares everyone."
Comments