Let us know about free updates
The author is a law professor at the University of Southern California Gould Law School and is the author of “High Wire: How China Regulates Large Technologies and Controls Its Economy.”
Last week, the US further tightened restrictions on semiconductor sales to China, causing a sharp drop in stock prices of US chipmakers Nvidia and AMD. In its quest for hegemony in the AI arms race, Washington appears to be determined to double export restrictions regardless of the cost to itself. But what happens if China isn’t trying to win?
American policymakers warn that those who control AI will gain critical economic, national security and military benefits. However, Beijing may have quietly concluded that being the closest second would be useful for its benefits.
The US dominates expensive, cutting-edge AI models. However, the rise of open source low-cost AI models like DeepSeek in China has already changed the demand for computing power in the global industry to respond to user queries from training AI models to using existing models. This is known as reasoning. Barclays estimates that inference will account for up to 70% of total AI computational demand by 2026.
This explains why the White House moved to limit Nvidia’s H20 chip sales. This is the most advanced chip that Chinese customers can buy from Nvidia and is optimized for inference.
However, this policy could help the Chinese government seek technical self-sufficiency and promote it at a lower cost. With Nvidia off the table, Chinese companies may use domestic alternatives, especially those offered by Huawei and Cambricon.
At a symposium between President Xi Jinping and Chinese tech entrepreneurs in February, Huawei founder Ren Zhengfei sat directly across from Xi.
Speculation helps lower the technical barriers for Chinese hardware manufacturers. According to a study by Deepseek, Huawei’s Ascend 910C chip already offers around 60% of the inference performance of Nvidia’s H100. In fact, Deepseek is reportedly already running some of its inference workloads on Huawei’s latest chip.
The US ban on tipping exceeds China. Towards the end of the Biden administration, the US has introduced the “AI Spreading Framework.” It divides the country into three layers and extends export restrictions to tier countries, including India, Malaysia, Singapore and the United Arab Emirates. This is expected to take place in mid-May unless the Trump administration withdraws the rules.
Tier 2 countries don’t sit still. Many harbor ambitious plans have built their own world-class data centers. In the face of our tipping restrictions, they may turn to China as an alternative source.
Therefore, to curb China’s progress, Washington may have yet another favor on Beijing by promoting both domestic and international demand for China’s chips.
Demand is also a powerful factor in innovation. Huawei currently leads a coalition of over 2,000 companies, working towards its goal of making China more than 70% self-sufficient across its semiconductor supply chain by 2028. Computing segments in AI data centers. Breakthroughs will help China overcome key bottlenecks in building large-scale AI computing infrastructures.
China is still behind the US in both AI hardware and software. That could be the case in the near future. However, this is not a barrier to countries promoting the AI agenda. You may be totally satisfied with the remaining second rest.
