U.S. Plans to Ban Chinese Optical Transceivers from AI Data Centers, Zhongji Innolight Takes the Biggest Hit
0xBroomberg
The FCC is considering a ban on Chinese-made optical transceivers in the US market. Zhongjixuchang (中际旭创), which holds roughly 27% of the global data-center optics market, stands to lose its largest revenue base — marking a new front in tech restrictions, from chips to data-center internals.
What are optical modules, and why are they suddenly a target?
Optical transceivers convert electrical signals to light and back — they are the components that let GPUs, servers, and switches talk over fiber at high speed. Without them, AI clusters cannot function.
As AI compute clusters scale up, demand for 800G, 1.6T and faster modules is surging. Chinese suppliers already dominate this market.
This means → US AI data centers rely heavily on Chinese supply chains for a core internal component, and that is exactly the dependency the FCC wants to sever.
How would the FCC enforce a ban?
According to Reuters, the FCC may follow the playbook it used for Chinese drones and routers: block new products from entering the market first, without immediately ripping out installed equipment.
In plain terms = the ban would not force operators to remove existing 800G modules. Instead, it would shut the door on upgrades — Chinese suppliers could keep selling current-generation products but could not ship the 1.6T systems future AI clusters will need.
US officials cite the Huawei precedent: waiting until equipment was deeply embedded in American networks made removal costly and slow. This time, the goal is to act before Chinese optics reach the same level of entrenchment in AI infrastructure.
Why is Zhongjixuchang the hardest hit?
Counterpoint Research estimates Zhongjixuchang (中际旭创) holds roughly 27% of the global data-center optical-module market. Over 90% of its revenue comes from outside China, with customers including Nvidia and Google.
In June this year, the US Department of Defense added the company to a list of firms allegedly linked to the Chinese military, intensifying scrutiny.
This means → a ban would not just cost Zhongjixuchang the US market. It could shake the foundation of its entire overseas business — other regions' customers may tighten sourcing in response.
Can alternative suppliers fill the gap?
US optical firms Lumentum, Coherent, and Applied Optoelectronics, along with Nokia, Ciena, and Cisco, could benefit from supply-chain reshuffling.
But these companies' production scale lags well behind Zhongjixuchang's, and a full replacement in the near term is unlikely.
This reflects a core tension: moving too fast on the ban would lengthen equipment-qualification cycles, tighten supply, and raise AI data-center build costs — directly conflicting with Washington's push to accelerate AI infrastructure expansion.
Where does this end up?
FCC officials aim to publish a formal proposal by the end of 2026, but the scope could still be narrowed or shelved.
The ban's final shape and pace of enforcement will determine whether the policy can balance security objectives against supply-chain stability.
In plain terms = ban too fast, and AI buildout hurts itself first; ban too slow, and the dependency only deepens — the timing window is the biggest variable of all.
Content is for reference only, not financial advice.