The United States is expanding its efforts to restrict China’s access to advanced computing by targeting not only semiconductor exports, but also the use of U.S.-made chips in data centers outside China. The policy addresses a long-known gap in export controls, while raising difficult questions about enforcement, costs for American companies and whether restrictions will accelerate China’s development of alternative AI technologies.
Cloud computing created a gap in chip controls
For a long time, Chinese companies did not necessarily need to smuggle advanced chips into China to use them. They could instead rent computing power from data centers in other countries, a practice that generally did not violate U.S. export regulations.
This was not a hidden loophole. Policymakers and researchers working on semiconductor restrictions were well aware of it. One reason the practice continued is that renting computing power can be more expensive and cumbersome than owning a chip, particularly for companies training large AI models over extended periods.
Some analysts also argued that allowing Chinese companies to use U.S. technology overseas could preserve their dependence on American suppliers rather than encouraging the development of domestic alternatives. Enforcing restrictions on cloud-based access is difficult as well, and could hurt American technology companies.
Washington expands technology controls
Washington is now attempting to address this issue. Monday’s announcement marks another significant expansion of U.S. efforts to limit China’s technological development.
In 2022, the U.S. government began blocking Chinese companies from acquiring certain U.S. semiconductors. Four years later, it is seeking to regulate how Chinese companies use those same chips, even when the hardware is not imported into China.
Whether that is a good idea depends partly on how effective the original restrictions have been. There is evidence that export controls have made it significantly harder for Chinese AI companies to obtain the computing power available to their U.S. competitors. However, measuring the precise impact of these policies has proven extremely difficult.
Export restrictions carry costs and risks
The strategy involves substantial risks, with clear winners and losers. Forcing U.S. semiconductor companies to cut off Chinese customers has already come at a cost to American businesses.
In research on previous U.S. export controls, New York Fed researchers found that affected U.S. suppliers were struggling to replace lost business in China with customers in other regions. They estimated that the restrictions eliminated $130 billion in combined market value for the companies involved, while also harming profitability and employment.
The regulations have also given China a strong incentive to develop AI systems without relying on U.S. technology. Huawei, for example, has developed a method for linking powerful chips to create larger and more capable megachip systems. Chinese AI lab DeepSeek has shown that more efficient training techniques can enable developers to build competitive models with fewer resources.
There is concern that the Trump administration may soon discover that it cannot fully control the outcome. The Chinese government is now seeking to regulate not only which chips Chinese companies can purchase from American companies, but also how they can use American technology anywhere in the world.
“I don’t know if I want to hold someone that responsible for what happens in someone’s sex life.”
This is an edition of Will Knight’s AI Lab Newsletter. Read previous newsletters here.
Source: www.wired.com


