The Trump administration is preparing new restrictions aimed at preventing Chinese companies from remotely accessing advanced U.S. AI chips located in overseas data centers. The proposed Commerce Department rule could target cloud-based access through third countries such as Thailand and Singapore.

Current U.S. export controls primarily restrict the physical shipment of advanced AI chips to China, leaving a significant loophole: Chinese companies can potentially access the same computing power remotely when chips are installed outside China.

Chinese technology companies including ByteDance, Alibaba and Tencent have reportedly accessed Nvidia-powered computing infrastructure through overseas data centers. The proposed rule would extend export-control principles from physical hardware to remote access to advanced computing capacity.

The move represents another shift in Washington’s rapidly evolving semiconductor strategy. Earlier policy changes had eased certain restrictions and moved some advanced-chip exports toward case-by-case reviews, while the latest proposal indicates renewed concern over China’s access to frontier AI infrastructure.

China’s accelerating AI capabilities are adding urgency. Advanced Chinese models have raised questions in Washington about whether existing restrictions can effectively limit access to the computing resources required to train increasingly powerful AI systems.

The larger challenge is that export controls can shift rather than eliminate access routes. Restricting direct chip shipments encourages companies to explore cloud infrastructure, intermediaries and alternative supply chains. Closing one loophole can therefore result in another pathway emerging.

There is also a broader strategic risk. Continued restrictions are accelerating China’s efforts to achieve semiconductor self-reliance. Huawei’s response to earlier U.S. restrictions demonstrates how technology controls can encourage domestic innovation, alternative architectures and indigenous supply chains rather than permanently stopping technological progress.

Ultimately, Washington faces a difficult balancing act between protecting advanced AI technology and maintaining a predictable global semiconductor market. Frequent policy changes create uncertainty for Nvidia, cloud providers, data-center operators and U.S. allies. A sustainable strategy will require more than closing individual loopholes—it will need coordinated export controls, cloud governance, international cooperation and a long-term semiconductor industrial strategy.