Jensen Huang on Chinese AI Models: Exceptionally Good, Should Be Used

Jensen Huang, CEO of U.S. chip giant NVIDIA (NVDA-US), stated that open-source artificial intelligence (AI) models developed in China are 'exceptionally good' and should be used by American enterprises. He also warned that restricting open-source models could weaken the United States' competitiveness in the AI field.

According to a report by U.S. media outlet Axios on April 22 (Wednesday), Huang said in an interview: 'These Chinese models are exceptionally good. Good open-source models should be used.' He emphasized that the U.S. has no reason to fear the advancement of Chinese open-source AI.

Huang's comments stand in stark contrast to the positions of the U.S. government and some American AI companies. He argued that open-source models will not push American firms out of the market, stating, 'This cannot happen—zero possibility.'

Huang believes the market has previously misjudged the impact of Chinese AI models. Regarding concerns sparked by Kimi K3, an AI model released by Chinese company Moonshot AI, he said Wall Street and U.S. policymakers 'have once again misunderstood the impact.'

In his view, cheaper and more open AI models will drive broader adoption of AI by businesses and individuals, ultimately increasing demand for NVIDIA's chips, data centers, and computing resources. 'Free AI should benefit hardware, benefit chips, benefit data centers,' he said.

The Kimi K3 model has recently caused a new wave of disruption in the AI industry, notable for its near-state-of-the-art performance, lower cost, and open-weight architecture. Some worry that such low-cost models could undermine the business rationale behind the U.S.'s large-scale investments in AI infrastructure.

Huang also questioned U.S. policies restricting Chinese AI models on national security grounds.

He pointed out that open models do not inherently contain 'backdoors,' and companies can manage risks through secure isolation environments. On the contrary, openness allows more researchers to identify vulnerabilities and strengthen defenses.

'If the world had only one model, one attack point, one source of failure, I believe the world would be far more vulnerable,' Huang said.