How China gets Nvidia chips despite US export restrictions

Nvidia CEO Jensen Huang's personal scandal has drawn attention to the challenge of how China obtains advanced AI microchips despite U.S. export restrictions. Nvidia designs powerful chips originally developed for video gaming visuals, called GPUs, which also have the capability to process vast amounts of data. These GPUs are used in training and running artificial intelligence applications, as seen with the launch of ChatGPT in 2022 and subsequent rapid investment in AI technology.

U.S. export controls prohibit the sale of Nvidia's high-end "Blackwell" and "Rubin" GPUs to China, citing security concerns, and this policy helps maintain global leadership in AI research and development. Despite these restrictions, Chinese AI startups like Moonshot AI, Z.ai, and DeepSeek have recently launched powerful AI models, demonstrating their ability to access top-tier Nvidia chips.

Chinese vendors reportedly obtain chips through smuggling, loopholes allowing them to rent computing power from data centers in third nations, or by training models on domestically produced chips. While U.S. advisers have accused Chinese AI labs of mimicking U.S. models, Nvidia has denied any evidence of chip smuggling. Chinese chip manufacturers, including Huawei, are rapidly improving their products and plan to launch next-generation AI chips sooner than originally anticipated.

However, Chinese AI companies may still face higher costs and longer training times when using domestic chips compared to Nvidia's solutions.