China transceiver ban would strain U.S. AI giants' infrastructure—Western makers can't fill gap in 1–2 years

Chinese optical module makers have built dominant supply-chain position in a critical piece of AI infrastructure. Innolight and Eoptolink (which generates 96% of its sales overseas) are the largest players, controlling roughly two-thirds of global transceiver supply. The Trump administration's proposed import ban aims to reduce dependence on Chinese hardware, a move that initially buoyed U.S. optical networking stocks—Coherent, Applied Optoelectronics, and Lumentum all surged on the news as investors bet on order redirection. However, Counterpoint Research and Raymond James have flagged a structural problem: Western suppliers cannot scale fast enough to fill the gap within the 12-to-24-month window that a rushed ban would impose. Coherent and Lumentum lack the cleanroom capacity, automated packaging infrastructure, and yield scale to absorb the displaced volume from Chinese competitors.

The consequence reaches far beyond optical component makers. Hyperscalers—AWS, Microsoft, Meta, and Google—collectively operate capital expenditure plans worth hundreds of billions of dollars, and many of those dollars are earmarked for GPU-dense AI superclusters that depend on optical transceivers for 400G, 800G, and 1.6T interconnects. A capacity crunch would translate directly into cost escalation and delayed deployments, causing lower utilization of expensive AI accelerators and rippling through multi-year infrastructure roadmaps. Counterpoint's Neil Shah explicitly warned that "sudden regulatory shifts risk creating hardware bottlenecks that could slow down deployment schedules for the world's largest cloud operators." The irony is sharp: a ban intended to protect U.S. AI competitiveness may instead harm the very cloud giants that drive it.