The Donald Trump administration is pursuing ways to restrict key China-made components used to build artificial intelligence (AI) data centers. After semiconductors and AI models, it is expanding the scope of regulation to data center network equipment, in a sign that the U.S.-China AI rivalry is spreading across the supply chain.

Cryptopolitan, a blockchain outlet, reported on Aug. 4 local time that the U.S. government is reviewing a plan to restrict the use in the United States of China-made optical transceivers for data centers.

Optical transceivers are key network devices that send and receive data via fiber optic cables inside data centers. They are considered essential components in large-scale data centers needed to train AI models, handling ultra-fast communication between servers.

The proposed curbs could be pursued in a form in which the U.S. Federal Communications Commission (FCC) restricts the inflow of new China-made optical transceivers into the U.S. market. It has not been finalized, but if implemented it is expected to have a significant impact on procurement strategies for AI data center operators in the United States.

The U.S. government is citing national security as the main reason for pushing the regulation. It judges that if equipment supplied by overseas companies is used in AI infrastructure, security risks such as data leaks, malware insertion and service disruption could arise.

Such concerns also intersect with recent security disputes surrounding China-made equipment. Researchers reported last year that they found a remotely accessible backdoor vulnerability in Unitree Robotics' robot dog Go1. U.S. regulators have also repeatedly raised security concerns about China-made power inverters used in data center power grids.

The market views China-based optical transceiver maker Zhongji Innolight as likely to be directly affected. The company was included in June on a U.S. Defense Department list of companies linked to the Chinese military and is said to hold about 27 percent of the global optical transceiver market for data centers.

U.S. companies Coherent and Lumentum are being mentioned as alternative suppliers that could benefit. However, both companies are assessed as lacking production capacity to meet all of the increased demand in the short term.

U.S. efforts to keep China in check are also expanding from hardware to AI software. The U.S. government is also reviewing regulation of China's open-weight AI models, but views in the industry are mixed.

Meta CEO Mark Zuckerberg (마크 저커버그) argued that the United States should focus on building better technology and more infrastructure rather than banning Chinese AI. Some U.S. AI startups are also concerned that blocking even low-cost Chinese AI models could instead weaken the competitiveness of U.S. developers.

China also signaled the possibility of countermeasures. The Chinese embassy in the United States urged Washington to stop regulations targeting Chinese companies and said it would take necessary response measures if its national interests are damaged.

The U.S.-China technology rivalry is thus showing signs of widening beyond semiconductors and AI model regulation to data center network equipment and the broader supply chain. Some forecast that the two countries' regulatory competition over AI infrastructure will intensify further.