FCC Weighs Ban on Chinese AI Transceivers Amid Supply Chain Paradox
A proposed import ban on critical data center hardware threatens to disrupt AI infrastructure while leaving the U.S. dependent on Chinese raw materials.
The Federal Communications Commission is drafting a rule to ban the import of new Chinese-made optical transceivers for U.S. data centers. The move is driven by national security concerns regarding potential malware and the risk of data exfiltration from critical AI infrastructure.
Optical transceivers are essential modules that convert electrical signals into light for fiber optic transmission, enabling the high-speed data movement required between GPUs in AI training clusters. The proposed ban targets a dominant sector of the market; Chinese manufacturers currently supply approximately 60% of the global market share for these modules. Among them, Zhongji Innolight alone holds roughly 27% of the global market share.
The Decoupling Context
This regulatory push is part of a broader strategic effort by the United States to decouple its technology stack from China, following similar restrictions on advanced semiconductors and AI chips. The tension has already sparked reciprocal actions from Beijing, which has pushed for the use of domestic AI chips in state-funded projects and ordered certain data centers to remove foreign-made hardware.
A Structural Paradox
While the FCC aims to reduce reliance on Chinese finished goods, the move reveals a deep structural dependency on Chinese raw materials. Western-made alternatives rely heavily on indium phosphide (InP) for high-speed optical chips. However, China controls approximately 70% of global indium production, meaning the very components intended to replace Chinese transceivers still depend on Chinese supply chains.
Industry Implications
If implemented, the ban could create a massive supply gap that Western manufacturers may struggle to fill rapidly, potentially slowing the deployment of AI infrastructure. This scarcity would likely drive up the cost of AI compute, which could translate into higher cloud pricing and API rates for end-users. The stakes are particularly high given the scale of investment in the sector; combined capital expenditures for major hyperscalers—including Alphabet, Amazon, Microsoft, and Meta—are projected to reach between $700 billion and $725 billion by 2026.
What's Next
Industry observers are now watching for the finalization of the FCC rule and whether the U.S. government will introduce parallel initiatives to secure non-Chinese sources of indium. Until a viable alternative for raw material procurement is established, the U.S. remains vulnerable to the same geopolitical leverage it is attempting to mitigate through the import ban.