Jensen Huang, Nvidia chief executive officer (CEO), publicly opposed moves by the U.S. government and domestic AI corporations seeking to keep Chinese open-source artificial intelligence (AI) models in check. He argued that Chinese AI models would not threaten U.S. AI dominance but instead expand the AI market and boost demand for semiconductors and data centers.
In an Axios interview published on the 22nd (local time), Huang said, "Chinese open-source AI models are excellent," and "good open-source models should be widely used."
Asked whether U.S. corporations should be allowed to use Chinese AI models, he answered, "Of course." This runs counter to the Trump administration, which is reviewing restrictions on the use of Chinese AI models, and to the positions of OpenAI and Anthropic, which have argued that China's AI technology threatens U.S. competitiveness.
He also dismissed concerns that Chinese AI would push out U.S. corporations, saying, "The chances of such a scenario unfolding are zero."
Huang predicted that the spread of open-source AI will grow the overall AI ecosystem. As more users first encounter AI through free models, some will later choose paid, closed models, and as a result the market size itself will increase, he said.
In particular, he expected that as AI service use increases, investment in the semiconductors and data centers needed to run it will also expand. He said, "Free AI can't help but be good for hardware, for semiconductors, and for data centers."
He also noted that Wall Street is overestimating the influence of Chinese AI models. As with the emergence of DeepSeek last year, he said the market is excessively worried about Moonshot AI's latest model "Kimi K3" as well.
He pushed back against claims that downloading Chinese AI models could pose security risks. Huang argued that because open-source models allow external researchers to verify source code and models to find vulnerabilities, openness can actually enhance safety.
In the same vein, he urged that Anthropic's AI model "Claude Mythos," which is available only to some users, be fully opened. He said, "It should be open to everyone," adding, "Tying up Anthropic does not align with America's interests."
The remarks came right after Scott Bessent, U.S. Minister of the Treasury, raised the possibility of sanctions on Chinese AI models. Minister Bessent said in a Fox Business interview that the U.S. is investigating whether Chinese AI models infringe U.S. intellectual property and, if necessary, could consider sanctions.
Minister Bessent claimed that a "watermark" from U.S. models is being found in Chinese AI models. In response, Huang said, "Distillation—that is, learning from another AI model—is a natural process in AI development," and emphasized that what matters is whether there was actual intellectual property infringement or wrongdoing, rather than the training method itself.