Nvidia CEO Jensen Huang is pushing back against US restrictions on Chinese artificial intelligence models, arguing that American companies should have access to them. His stance comes as South Korea's president prepares for high-stakes meetings with major tech leaders to solidify the country's role in the global AI supply chain.

Why Is Jensen Huang Defending Chinese AI Models?

Huang believes the market has fundamentally misunderstood the impact of affordable Chinese AI models like Kimi K3, the latest offering from startup Moonshot AI. Rather than reducing demand for computing hardware, he argues that cheaper and more accessible AI models will actually encourage wider adoption, driving higher demand for Nvidia's chips, data centers, and computing infrastructure.

The Nvidia chief dismissed security concerns about downloading Chinese open-weight models, which are AI systems that developers can download, customize, and run on their own infrastructure. He explained that organizations can deploy these models within secure environments and that open models actually improve security because researchers can inspect them, identify weaknesses, and build stronger safeguards.

"American companies should 'absolutely' be allowed to use Chinese AI models," Huang told Axios, arguing that banning them would do more harm than good.

Jensen Huang, CEO at Nvidia

Huang's position has gained support from Tesla CEO Elon Musk and nearly 200 Silicon Valley companies, including startup accelerator Y Combinator and encrypted email provider Proton. These companies have urged the Trump administration not to restrict access to Chinese open-weight AI models, warning that broad restrictions would hurt US innovation more than China's.

What's Driving the US Government's Concerns About Chinese AI?

The White House Office of Science and Technology Policy has accused Moonshot AI of intellectual property theft. Officials claim the company's Kimi K3 model was built using technology distilled from Anthropic's Claude Fable model, a technique where smaller, more efficient models are created by learning from larger ones.

White House officials alleged that Moonshot AI developed a "sophisticated internal platform" to conduct large-scale distillation of US AI models while attempting to avoid detection. The company reportedly acquired Nvidia GB300-equipped servers and accessed additional GB300 systems in Thailand to train its models.

While acknowledging that AI distillation is a common and legitimate technique, US Treasury Secretary Scott Bessent said the administration is investigating whether Chinese AI companies improperly distilled American AI models. Companies found to have stolen US intellectual property could face sanctions.

How South Korea Is Positioning Itself in the AI Infrastructure Race

South Korean President Lee Jae-myung will visit San Francisco from July 24 to 25 to attend an AI Summit where he will hold intensive meetings with major tech leaders. The delegation includes Samsung Electronics Chairman Jay Y. Lee, SK Group Chairman Chey Tae-won, Hyundai Motor Chairman Euisun Chung, and NAVER founder Hae-jin Lee.

The summit represents a strategic shift in how countries compete in artificial intelligence. Rather than focusing solely on developing the best AI models, the competition is now centered on controlling the computing power infrastructure that makes advanced AI possible.

• High Bandwidth Memory Supply: Samsung Electronics, SK Hynix, and Micron have all entered the HBM4 supply chain for Nvidia's next-generation Vera Rubin platform. As Vera Rubin enters mass production, Nvidia's requirements for HBM4 capacity, supply pacing, and product performance continue to rise, making this a critical focus area for South Korea's memory chipmakers.
• Foundry Partnerships: Nvidia CEO Jensen Huang confirmed that Samsung is manufacturing Groq 3 LPU inference chips for Nvidia using a 4-nanometer process. These chips are expected to enter mass production and launch in the third quarter, representing a significant manufacturing opportunity for South Korea.
• Data Center Infrastructure: NAVER announced an expansion of cooperation with Nvidia in June, planning to build AI infrastructure based on Nvidia's DSX platform starting with the Sejong GAK Data Center. The initial scale is 55 megawatts, with a long-term goal of expanding to the gigawatt level.

Analysts believe that South Korea's personal diplomatic push reflects a recognition that AI competition is extending from model capabilities to a comprehensive contest over computing power infrastructure systems. By leveraging its existing advantages in memory chips and foundry capacity, South Korea is positioning itself to secure dominance in the global AI supply chain.

The broader context reveals a fundamental tension in US technology policy. While government officials worry about Chinese AI companies closing the gap with American firms, industry leaders like Huang argue that restricting access to affordable AI tools will ultimately harm American innovation and competitiveness. Meanwhile, countries like South Korea are strategically positioning themselves to profit from the infrastructure that powers all of these competing AI systems, regardless of where the models originate.