China Won’t Get Addicted to America’s Chips

As Rush Doshi and Chris McGuire note, selling chips to China would liquidate U.S. leverage, turning scarcity into stockpiles and handing Beijing the parts and know-how to close the innovation gap (Letters, Sept. 2). The notion that we can get China addicted to our chips ignores Communist Party doctrine. Beijing’s strategy seeks to develop indigenous and controllable tech to achieve self-reliance from the West. To that end, the state gate-keeps market access, data and content through licensing, cybersecurity and procurement rules, as well as other conditional, revocable permissions.

The recent H20 episode illustrates the point. After Washington licensed sales, Beijing told firms to make a priority of domestic alternatives. Companies may exploit brief windows to stockpile—much as they bought A800/H800s in 2022-23 before that loophole closed—but the policy direction is unchanged. Beijing will take what it needs today to reach self-reliance tomorrow.

Critics of export controls are right that some contraband chips will leak. But that proves that controls bite. What reaches the market are small, warranty-free, irregular lots—not the steady, uniform, high-volume supply that large-scale training requires. When enforced, controls raise costs, stretch timelines and cap scale by constraining the chokepoints that matter: training-class accelerators, HBM, EDA tools and lithography.

Some suggest that CUDA, Nvidia’s software for programming chips, would “tether” China to the American AI ecosystem. Yet authorizing supported CUDA deployments would replace today’s gray-market trickle with stable, vendor-backed systems that train Chinese engineers and accelerate domestic substitution. Beijing could then shut down access at will, trading our leverage for its own learning.

Craig Singleton is a senior China fellow at the Foundation for Defense of Democracies and a former U.S. diplomat. X: @CraigMSingleton