Microsoft, which currently has 35 data-center projects under construction and another 134 slated for the future, has joined a small coalition of U.S. technology firms to lobby the Trump administration against a proposed import restriction on Chinese-made optical transceivers. The companies argue that the rule would create a supply-chain bottleneck that could cripple the nation’s race to deploy AI-powered compute capacity.
The concern centers on the fact that modern data-center architectures rely on thousands of these tiny modules to translate electrical signals from servers into light that travels over fiber-optic cables. Without them, even the most advanced AI chips would sit idle, unable to communicate with each other at the required speeds.
Industry leaders meet White House science office and National Security Council
In September, Microsoft and a handful of peers met with officials from the White House Office of Science and Technology Policy and the National Security Council. The briefing highlighted how a ban on new Chinese models could disrupt the rollout of next-generation infrastructure. Representatives from the Information Technology Industry Council (ITI) echoed the same message, noting that the United States risks losing its competitive edge in artificial intelligence if critical components become scarce.
John Miller, ITI’s general counsel, told officials that the proposed rule “really threatens to arguably grind the data-center rollout to a halt.” He added that optical transceivers are the connective tissue between GPUs, switches, and routers; without them, the expensive AI processors already procured by cloud providers would remain “un-utilized.”
FCC’s draft rule and the 3.2-terabit threshold
The Federal Communications Commission is drafting an equipment-authorization rule that would bar U.S. imports of new Chinese optical transceiver models deemed a security risk. The agency’s tool works by denying FCC approval to devices that contain logic-bearing components from entities on its “Covered List.” While the leading Chinese manufacturers—Zhongji Innolight and Eoptolink—are not yet on that list, the FCC could create a new designation specifically for transceivers.
One version of the proposal would limit the ban to transceivers capable of 3.2 terabits per second (Tbps). Morgan Stanley analysts have noted that products meeting this speed are not expected to reach volume production until 2028, with widespread deployment projected for 2029. By setting the threshold above today’s 800 gigabit (G) and 1.6-terabit (T) optics, regulators might avoid immediate disruption while still addressing long-term security concerns.
Supply-chain dynamics and the risk to U.S. AI leadership
According to Counterpoint Research, Chinese firms command roughly two-thirds of the global optical transceiver market. Eoptolink and Zhongji Innolight dominate the segment, and both supply a substantial share of the components used by U.S. hyperscale operators. Although domestic players such as Coherent and Lumentum produce competing parts, they lack the scale to replace Chinese volume quickly.
Beyond market share, security arguments focus on the firmware and management interfaces embedded in modern transceivers. Analysts like Irina Tsukerman argue that these software layers could be leveraged for data exfiltration or malware insertion, making the devices more than “comparatively dumb” signal converters.
Legislators have also entered the fray. A bipartisan group introduced a bill in September that would prohibit the federal government from installing Chinese-made transceivers in any sensitive system. The move aligns with broader U.S. efforts to restrict Chinese semiconductor technology, especially as the current trade truce with Beijing approaches its January 10 deadline.
Microsoft has declined to comment on its lobbying activities, and the White House, FCC, and Chinese embassy have not responded to requests for statements. Nonetheless, the company’s position underscores a broader industry fear: that a swift, blanket ban could halt the construction of AI-ready data centers and cede technological momentum to rivals.
