Washington increasingly frames artificial intelligence as a single decisive race toward general intelligence. China’s strategy points elsewhere. The danger is not building AI, but locking policy into a worst case narrative that turns uncertainty into irreversible escalation.
Western headlines claim China is bypassing chip export controls. A close reading of Chinese and Taiwanese sources tells a different story: slow progress, rising costs, and no proven evidence of illicit upgrades. This analysis separates verifiable fact from allegation and explains what China’s DUV based strategy actually achieves.
Donald Trump has reopened the door for Nvidia’s H200 sales to approved customers in China. Beijing’s response is not to celebrate but to ration access, shield Huawei and deepen its own AI hardware stack. This article follows on from our investigation into offshore Chinese model training and explains how both Washington and Beijing now run export style controls on the same chip.
China’s artificial intelligence giants are not only dodging United States export controls. They are also navigating Beijing’s clampdown on Nvidia. New rules that bar fresh Nvidia deployments in Chinese data centres are pushing Alibaba, ByteDance and DeepSeek to rent GPU farms in Singapore and Malaysia, even as they are forced to build a parallel stack on Huawei and other domestic chips at home.