reuters.com web signal

FCC Drafts Import Ban on Chinese Data-Center Transceivers

TL;DR

  • The FCC is reportedly drafting a rule to bar imports of new Chinese optical transceivers over risks of data theft, malware, and service disruption at AI data centers.
  • Zhongji Innolight, estimated to control around 27% of the global data centre transceiver market, was added to the Pentagon's Chinese military-backed companies list in June.
  • US alternatives Coherent and Lumentum could benefit, though analysts say they currently lack the manufacturing scale to fully replace Chinese producers.

The FCC is reportedly drafting a rule that would bar imports of new Chinese optical transceivers, the small modules that let data travel over fiber-optic cables at the speed of light inside data centers. According to Reuters, the move aims to prevent Chinese firms from stealing data, installing malware or disrupting service at U.S. facilities that now house the chips training and running AI models. Officials hope to publish it this year, when it would take effect, though the FCC could still modify or shelve the restriction.

The company most exposed is Zhongji Innolight, which industry analysts estimate controls around 27% of the global data centre transceiver market, and which was added to the Pentagon's list of alleged Chinese military-backed companies in June. Innolight pushed back in a June disclosure, saying the designation was inconsistent with the facts and that it is neither a Chinese military company nor a military-civil fusion enterprise. Take the security rationale as reported, not settled; the piece is sourced anonymously and does not point to a specific breach, only to the class of risk the rule is meant to head off.

The read-through for hyperscalers is that a hardware line-item most buyers barely think about becomes a scarcity item. Optical modules are the plumbing of AI training clusters, and re-qualifying suppliers is not instant. American names like Coherent and Lumentum could pick up demand if cloud providers shift away from Chinese suppliers, though analysts note they currently lack the manufacturing scale to fully replace Chinese producers, which is the kind of gap that tends to show up as build-out delays and tighter margins well before it shows up in a policy filing.

What the reporting doesn't give you is a timeline for when the rule text actually lands, how narrowly 'new models' will be defined, or whether existing installed base will be grandfathered. Those details are what decide whether this is a cliff or a slope for the AI buildout schedule, and they are the things worth watching between now and publication.