As Washington grapples with the rapid advancement of Chinese artificial intelligence, Nvidia CEO Jensen Huang is urging U.S. policymakers to embrace rather than restrict open-source models. In an interview with Axios, Huang described models developed by firms like Moonshot AI, DeepSeek, and Alibaba as "excellent," arguing that they should be utilized by American companies.
The Rise of High-Capability, Low-Cost Models
The debate has intensified following the July 16 release of Moonshot AI’s Kimi K3. The model has shown the ability to compete in coding benchmarks with Anthropic’s Fable 5, one of the most advanced U.S. models. However, the price disparity is stark: Kimi K3 costs $15 per 1 million output tokens, compared to $50 for Fable 5. Other Chinese alternatives, such as DeepSeek V4, are priced as low as 87 cents for the same output.
Security Concerns vs. Customization
The White House is reportedly considering executive action to impose conditions on U.S. firms using Chinese AI, citing fears of surveillance and threats to the viability of domestic AI companies. Huang dismissed the notion that these models serve as a "backdoor" for the Chinese government, calling it a "misconception." He emphasized that because these models are open-source, the security guardrails can be fully customized by the organizations deploying them.
Nvidia’s Strategic Interest
Huang’s stance is closely tied to the demand for hardware. He argued that the proliferation of high-quality AI, even from international competitors, fuels the need for more data centers and computing power. "Whenever there’s more use, you’ll have to sell a lot more NVIDIA computers," Huang noted, suggesting that the diffusion of technology across industries ultimately benefits the broader infrastructure market.
