The Trump administration's Federal Communications Commission is drafting a ban on imports of new Chinese-made optical transceivers — the pluggable fiber-optic modules that move data at the speed of light inside AI data centers — in what officials describe as a national security measure to harden America's AI infrastructure. Four people familiar with plans told Reuters the proposal would prohibit new models from Chinese manufacturers and could take effect before year-end 2026, though the FCC could still modify or abandon it.

There is a structural contradiction at the center of this policy. The Western manufacturers — Lumentum and Coherent — that the FCC intends to empower as replacements for Chinese suppliers depend on indium phosphide, a compound semiconductor material, to build the high-speed lasers inside every competitive transceiver they make. China controls approximately 70% of global refined indium production and placed indium phosphide on its export control list in February 2025, causing Chinese indium exports to fall roughly 72% year-over-year by late 2025, per Reuters reporting. The price of a 6-inch indium phosphide wafer has risen roughly 250% since those controls began, reaching approximately $5,000.

If China tightens those controls further in retaliation for the transceiver ban — a response its government has explicitly refused to rule out — the ban's intended beneficiaries cannot scale fast enough to replace what they would be removing.

What Optical Transceivers Are, and Why AI Clusters Can't Run Without Them

Optical transceivers are small, pluggable devices that convert electrical signals into pulses of light and back again, enabling data to traverse fiber-optic cables at extraordinary speeds and with far lower power consumption than copper wiring. Inside a modern AI data center, they are everywhere: linking GPU servers to network switches, connecting racks across a facility floor, and carrying the massive data flows that training large AI models demands.

The enabling component at every speed tier above 100 gigabits per second is an electro-absorption modulated laser, or EML — a precision device that combines a laser source and high-speed modulator on a single indium phosphide chip. According to McKinsey's optical supply chain analysis, fewer than five companies worldwide manufacture EMLs at commercial scale. At 200 gigabits per lane — the signaling speed required for next-generation 1.6-terabit modules — passive copper cables fail to reach across even a single server rack without signal degradation. As prior TechTimes reporting confirms, optical interconnects built on indium phosphide lasers are not optional for AI infrastructure at current scales; they are a prerequisite.

A single GB200 GPU cluster containing 576 processors requires tens of thousands of individual 800G optical modules. As clusters have expanded from thousands to hundreds of thousands of GPUs, the optical interconnect layer has emerged as a binding constraint sitting alongside raw compute. The dominant speed tier today is 800 gigabits per second (800G), with 1.6-terabit modules already entering production.

Chinese Suppliers Dominate the Market the Ban Would Upend

Chinese manufacturers hold approximately 60% of global 800G market share, with pricing running 20 to 25% below Western incumbents — a cost advantage rooted in aggressive manufacturing scale and deep integration with NVIDIA's qualification processes.

The two companies most directly affected by the proposed ban are Zhongji Innolight and Eoptolink, which together supply the majority of NVIDIA's 800G module demand.

Innolight, based in Longkou, Shandong, posted RMB 19.5 billion (approximately $2.9 billion USD) in revenue in the first quarter of 2026 alone — a 192% year-on-year increase — with net profit of RMB 5.74 billion (approximately $844 million USD), up 262%. That single quarter's profit exceeded the company's entire 2024 earnings. Innolight holds a 27% share of the global data center transceiver market, according to Counterpoint Research, and was added to the Pentagon's list of alleged Chinese military-backed companies in June 2026. Its customer list spans American big tech: Alphabet accounts for 22% of Innolight's revenue, Amazon for 11%, and Meta for 6.4%, with more than 90% of its revenue coming from outside China.

Eoptolink, headquartered in Chengdu, Sichuan, posted Q1 2026 quarterly revenue of approximately $1.2 billion, representing 106% year-over-year growth, with net margins of approximately 33% — among the highest in the sector.

Both companies have moved to reduce their tariff exposure by shifting capacity to Thailand. Whether production from those facilities would fall inside or outside the scope of an FCC restriction depends entirely on how the agency defines a "Chinese optical transceiver" — a question whose answer remains unresolved in the draft. Bloomberg Intelligence analyst Sean Chen noted that Chinese companies may also be able to "circumvent FCC restrictions via Southeast Asia, depending on how the FCC defines a Chinese optical transceiver."

The Security Rationale: Real Risk, No Confirmed Incident

Officials say the proposed restriction aims to prevent Chinese firms from stealing data, installing malware, or disrupting service at US data centers, which house the chips used to train and run AI models.

Experts say the concern is architecturally credible even without a confirmed incident. Modern optical transceivers contain firmware, onboard memory, and management interfaces; geopolitical analyst Irina Tsukerman, cited by Network World, noted that those components could theoretically "influence how traffic is monitored and managed throughout the data center," with risks potentially extending to "compromised firmware updates, manipulation of diagnostic information, disruption of maintenance support... or in the worst case scenario, interference with critical infrastructure during periods of heightened political tension."

"Transceivers definitely pose a risk," said Divyansh Kaushik, an AI policy expert at Beacon Global Strategies. "Data center supply chain security must be locked in from the get-go as buildout scales up," he added.

No confirmed data-access incident involving Innolight or Eoptolink has been reported. Optical modules are physical hardware rather than data-collection software, and that distinction matters for risk assessment. One legal condition applies to both companies regardless of product design, however: China's 2017 National Intelligence Law, Article 7, obligates all Chinese organizations to "support, assist, and cooperate with national intelligence work." That obligation is not contingent on the company's stated privacy policy or the physical location of its factories.

A Pattern of Escalating FCC Action

The proposed transceiver ban follows an accelerating series of FCC interventions targeting Chinese technology. The same regulatory framework shut Huawei out of US telecom networks beginning in 2021, grounded DJI's drone business in December 2025, and blocked new Chinese-made consumer routers in March 2026. Just one week before the transceiver ban was reported, the FCC added Chinese robots and power inverters — citing documented backdoors — to its Covered List.

The optical transceiver proposal represents a strategic evolution in the approach: rather than reacting to Chinese technology already embedded in US networks, officials are trying to stop next-generation components from gaining a foothold before they become as difficult to dislodge as Huawei's telecom gear once was. The FCC Covered List framework, established under the Secure and Trusted Communications Networks Act, does not require full public notice-and-comment rulemaking, allowing the agency to act quickly on a White House interagency national security determination.

A House Select Committee China report released Tuesday found that regulatory gaps had allowed three Chinese telecommunications companies barred from directly connecting to US networks to retain links to American internet infrastructure — and called for the FCC to extend network restrictions further.

Beijing's Hidden Leverage: Indium Phosphide

The ban's structural vulnerability runs deeper than a supply shortfall. China's February 2025 export restrictions on indium phosphide — the compound semiconductor on which all competitive optical lasers are built — have already functioned as a trade weapon, pushing the price of a 6-inch indium phosphide wafer from roughly $1,400 to $5,000 and causing AXT, one of the world's largest indium phosphide substrate producers and a key supplier to Coherent, to report over $60 million order backlog as of the first quarter of 2026. Coherent's CEO Jim Anderson traveled to Beijing alongside a US business delegation accompanying President Trump in May 2026, partly to raise the issue of delayed InP export licenses with Chinese authorities directly — a measure of how seriously the Western manufacturing sector takes the constraint.

"InP is one of several supply chain bottlenecks gating AI data centre buildouts," said Konrad Wang, a research analyst at SemiAnalysis.

The chokepoint is structural, not cyclical: indium phosphide cannot be substituted in high-speed optical applications. Silicon cannot achieve the electron mobility InP provides at signaling speeds above 100 gigabits per lane. Even with substantial capital investment — NVIDIA committed $2 billion each to Lumentum and Coherent in March 2026, and the CHIPS Act proposed a further $50 million grant for Coherent's Texas facility — the wafers those facilities require still flow primarily through China's export permit system.

Lumentum CEO Michael Hurlston has warned that indium phosphide shortage exceeds memory — a shortage more severe than the DRAM crunch of early 2026. An easing occurred in May 2026 when China released a batch of indium phosphide substrate export permits, but AXT backlog data confirmed that relief is selective and politically contingent, not structural.

The Supply Gap Problem

Even without an indium constraint, American manufacturers currently lack the manufacturing scale to fill the void a ban would create.

LightCounting April 2026 forecast puts current transceiver demand approximately 30% above supply, and its most recent July 2026 update projects Ethernet transceiver sales growing 73% in 2026 alone. That gap would widen sharply if Chinese supply were cut off.

"If you remove one provider from the equation, it's not as if the others have capacity to fulfil the increase in demand, so it's not simply a question of added cost, it's a question of placing a ceiling on capacity and growth," said Nader Henein, a Gartner vice president analyst.

Applied Optoelectronics completed its first volume 800G shipment to a large hyperscale customer in recent months and guided full-year 2026 revenue to potentially exceed $1 billion, against roughly $455 million in all of 2025. US manufacturing capacity stands at approximately 100,000 800G units per month — a figure that covers a fraction of current market demand.

One Foundation for American Innovation analysis concluded that Coherent and Lumentum sell competitive technology but currently lack the manufacturing scale to replace Chinese vendors. Analysts have suggested a middle path: rather than banning all Chinese 800G modules immediately, the FCC could restrict only next-generation 1.6-terabit module imports from Chinese vendors, allowing Western manufacturers time to scale without disrupting installed 800G infrastructure.

Market Reaction

Financial markets reacted swiftly to the Reuters report on August 4, 2026. Shares of US transceiver manufacturers viewed as potential beneficiaries surged: Lumentum gained 7%, Coherent rose 11%, and Applied Optoelectronics jumped 18%. Marvell Technology shares, a US optical-communications company, advanced more than 8%. On the other side, Innolight Shenzhen shares fell as much as 14% before paring the decline. Shares of hyperscaler operators who would face higher procurement costs also moved lower, with Amazon falling more than 1.7%.

Beijing Pushes Back

China's response came swiftly and at the highest diplomatic level. The Chinese embassy in Washington urged the United States to "heed the objective and rational voices of the business communities in both countries" and to "stop smearing Chinese companies and threatening them with sanctions," adding that "China will take all necessary measures in response to any action that causes material harm to its interests."

China's state-owned People's Daily described the proposed ban as part of an effort to "suppress Chinese technology and contain the country's industrial development," and warned that the move "runs counter to stabilisation and improvement in China-US relations," referencing the May summit between Xi Jinping and Donald Trump. "If the US side is bent on going its own way, China will take resolute countermeasures," the newspaper added.

The timing is diplomatically sensitive: the ban proposal came as officials on both sides were working to stabilize relations ahead of a Xi Jinping US September visit expected in September 2026.

Analysts noted that China's most potent available retaliatory lever is not a counter-tariff but the indium phosphide export permit system — the same mechanism it deployed in February 2025, before the transceiver ban proposal existed.

What Comes Next

The proposed rule remains a draft. The FCC could still alter or abandon it. If it moves forward, the most likely pathway is through FCC Covered List mechanism, which does not require a full notice-and-comment rulemaking and allows non-Chinese suppliers time to be exempted before restrictions take effect.

For the broader AI infrastructure buildout, the timing is acutely consequential. With GPU clusters continuing to scale and optical interconnect becoming a binding constraint on AI training throughput, the components that move data between chips are no longer a background procurement detail. They are now a geopolitical instrument — and the raw material on which the alternative supply chain depends is still controlled by the country whose products the ban would remove.

---

Frequently Asked Questions

What does the FCC's proposed ban on Chinese optical transceivers actually do?

The Federal Communications Commission is drafting a rule that would bar new models of Chinese-made optical transceivers — the pluggable devices that convert electrical signals to light and back, enabling data to travel through fiber-optic cables inside AI data centers — from receiving the equipment authorization required to be imported, marketed, or sold in the United States. The ban would apply to new models; existing authorized hardware would not be retroactively affected. The rule is a draft and has not yet been published or put into effect. Officials hope to publish it before year-end 2026, but the FCC could still modify or abandon it.

Why can't US manufacturers simply replace Chinese transceivers right away?

The core constraint is not assembly capacity — it is the raw material that powers the lasers inside every competitive high-speed transceiver. Indium phosphide, a compound semiconductor with no viable substitute at signaling speeds above 100 gigabits per lane, is produced approximately 70% in China. China placed indium phosphide on its export control list in February 2025, causing prices to rise 250% and creating documented supply backlogs at US substrate producers like AXT. Coherent and Lumentum — the two Western companies the ban is meant to empower — depend on indium phosphide to scale production. If China tightens those controls further in response to the transceiver ban, the Western production ramp the FCC is counting on cannot materialize fast enough to cover the gap.

Is there actual evidence that Chinese optical transceivers have been used to steal data or install malware?

No confirmed incident involving Chinese optical transceivers has been reported. The FCC's rationale is prospective: modern transceivers contain firmware, onboard memory, and management interfaces that represent a theoretical attack surface. The national security concern is also structural — China's National Intelligence Law of 2017, Article 7, legally requires all Chinese organizations to cooperate with state intelligence requests. That obligation is not contingent on the company's stated privacy policy or where its products are physically assembled. The risk is architectural and legally grounded, even without a documented incident.

What would a reader who might buy or deploy AI infrastructure need to know right now?

Enterprise and hyperscale buyers should be aware that the ban, if finalized, would apply only to new transceiver models — existing inventory and already-authorized hardware can still be deployed. In the near term, stocking up on current-model Chinese transceivers before any restriction takes effect may be rational. Longer term, the supply picture for Western alternatives is constrained by both manufacturing capacity and the indium phosphide raw material shortage; any transition will take longer and cost more than a simple supplier switch. The ban's final scope — specifically, how the FCC defines a "Chinese optical transceiver" and whether third-country-assembled units qualify — will determine how much Chinese production capacity survives the restriction intact.