An exclusive report from Reuters directly stirred up the optical communication sectors in the U.S. and China. On August 4th, AAOI surged 19% in U.S. trading, while Coherent and Lumentum saw intraday gains of 15% each. Even Corning, an upstream fiber optic material supplier, recorded a gain of nearly 8%.
On August 5th, Hong Kong-listed optical communication concept stocks generally declined. InnoLight (03308.HK) fell 7% at the market open, and Cambridge Technology (06166.HK) dropped 5%.
How should we correctly view this sudden event?
01
First, let's examine the core of this rumor: The FCC is drafting a new regulation, planning to ban the import of new Chinese-made optical modules into U.S. data centers. The target is to announce it this year and have it take effect immediately. It primarily targets high-speed optical transceivers supporting AI computing power, namely the main AI hardware products of this generation like 800G and 1.6T.
There is a key point of ambiguity: The draft does not clearly define what constitutes "new"—whether existing models, inventory already cleared customs, and optical components/parts are included in the ban's scope are all unknown.
And these details will directly determine the ultimate magnitude of the impact.
According to details in the Reuters report, the FCC will ban the import of all new transceiver models but plans to exempt many non-Chinese suppliers.
The design of this "exemption for non-Chinese suppliers" means the ban precisely targets optical module products from Chinese brands, not all transceivers manufactured within China. This distinction determines whether products from Chinese brands manufactured in Southeast Asian factories are similarly restricted.
However, it's worth noting that the FCC voted on July 22, 2026, to pass a proposal expanding the scope of the ban, prohibiting the import of equipment containing logic hardware components produced by "covered list entities."
This means that even if the optical module itself is not "Assembled in China," as long as it contains core logic chips produced by Chinese companies, it could also be subject to restrictions.
There are still too many unknown details, leading to significant uncertainty.
The pre-market surge in U.S.-based manufacturers essentially trades on order substitution expectations. The logic is simple: If new-generation Chinese optical modules cannot enter the U.S. market, North American cloud companies' AI computing capital expenditures will have to shift towards the domestic supply chain.
Applied Optoelectronics (AAOI.US), as a company directly involved in high-speed datacom optical modules, is the most direct substitute, hence the highest elasticity.
Lumentum (LITE.US) and Coherent (COHR.US), holding laser and high-speed optical chip technologies, have had their capacity secured by NVIDIA through strategic investments. Being both upstream chip suppliers and module OEMs, they were naturally heavily favored by capital.
Corning (GLW.US) benefits more from the increased demand for fiber optic and optical interconnect materials driven by computing power construction, making it an indirect beneficiary within the industry chain.
However, an unavoidable hard reality is that the effective capacity of the current few U.S.-based manufacturers could not fill the gap vacated by Chinese manufacturers within six months.
Although U.S. manufacturers have clear expansion plans, the capacity gap remains enormous.
Global shipments of 1.6T modules in 2025 are projected to reach 4.2 million units, coupled with massive demand for 800G modules. The combined capacity of the non-Chinese camp (Lumentum ~100,000 units/month, AOI plans for 500,000 units/month, Fabrinet's Thailand factory expanding) falls far short of filling the void left if Chinese manufacturers are excluded.
Orders for some Chinese optical module manufacturers are already scheduled until 2028, indicating supply itself is in an extremely tight state. Meanwhile, domestic leading companies occupy over 70% of global high-speed optical module capacity, with InnoLight alone holding about a 27% share in the U.S. datacom optical module market.
Policy can close the import door with a single document, but capacity ramp-up, yield rate refinement, and major customer validation all take time—it's not something that can be filled in just a few months.
This also represents the biggest risk for a subsequent stock price correction: Once the market realizes the realization cycle for benefits is far longer than imagined, or if the draft incorporates many exemption clauses, the optimistic premium could quickly be given back.
In fact, this ban is not an isolated incident but a continuation and escalation of the Trump administration's technology restriction policies towards China.
On July 28th, the FCC had already announced a ban on importing Chinese-made humanoid robots and power inverters, citing reasons of protecting national security in U.S. AI infrastructure construction.
Previously, the FCC had also banned the import of foreign-manufactured consumer-grade routers.
Actually, the U.S. has been taking various small actions to set restrictions on China under various pretexts; it's just that they are often overlooked by everyone.
Additionally, here are the latest interpretations from Morgan Stanley and Citi:
Morgan Stanley's core judgment is that this news is favorable for the U.S. optical component supply chain. Regarding the benefit logic, Morgan Stanley pointed out that InnoLight (03308.HK) and Eoptolink (300502.SZ) together account for about 50% of the global optical transceiver market. Under this landscape, Coherent (COHR.US) becomes the clearest, large-scale beneficiary due to its vertical integration capabilities. Applied Optoelectronics (AAOI.US) and Fabrinet (FN.US) are also expected to capture some incremental demand. While Lumentum (LITE.US) has relatively small direct optical module exposure, it supplies lasers to multiple global module manufacturers. Restrictions on Chinese suppliers will reduce the likelihood of easing the current tight supply situation, meaning the tight supply state for EML may persist longer, thus benefiting indirectly.
Regarding feasibility, Morgan Stanley's stance is relatively cautious. The report bluntly states that the current capacity of non-Chinese manufacturers is insufficient to meet the demand driven by AI capital expenditures. Notably, the CEOs of both LITE and COHR have publicly advocated for such restrictions, arguing that it's unfair as U.S. manufacturers generally cannot supply Chinese cloud companies, while Chinese manufacturers can supply the U.S. Additionally, over the past two years, most cloud companies have anticipated the possibility of such policies, and many have already completed certification of alternative suppliers, even though their current capacity cannot fully meet demand. The report also mentions that such policies can help alleviate market concerns about the "profit margin ceiling" for optical module manufacturers in the short term.
The biggest risk point lies with Indium Phosphide (InP) substrates. Morgan Stanley points out that InP substrate supply mainly comes from Chinese manufacturers (like AXT Inc (AXTI.US))—LITE just announced a new supply agreement last week, and COHR's CEO visited China months ago as part of Trump's visit. Sustaining access to InP substrates will be crucial to meet the surging demand. The report also suggests a possible solution: Chinese cloud companies committing to purchase U.S. components in exchange for some level of compromise.
Citi's view can be summarized as follows: This matter will not be a simple, clear-cut "one-size-fits-all" rule, for four reasons: Seven of the world's top ten optical module manufacturers are Chinese companies, supplying over 50% of high-speed optical modules to major U.S. CSPs; the current AI optical component supply chain itself is tight, and leading Chinese manufacturers still hold key advantages in component supply security and advanced product R&D; Chinese module manufacturers have a clear cost advantage, supporting the capital expenditure efficiency of CSPs; leading Chinese optical module manufacturers have all established overseas factories. Based on these market realities, Citi believes the final rule will likely contain certain exemption clauses, and overseas capacity expansion will remain a core strategy for Chinese suppliers.
02
If Chinese optical modules are excluded from the U.S. market, leading domestic optical module manufacturers have varying degrees of exposure:
Eoptolink (300502.SZ) has the highest dependence on North American customers. In 2024, 87% of Eoptolink's revenue came from overseas markets, making it, alongside InnoLight, one of the most aggressive Chinese optical module exporters. If the ban is implemented, the pressure from new product entry barriers would be most direct for Eoptolink, whose 2025 revenue is $3.5 billion. InnoLight (03308.HK) also has a high proportion of U.S. revenue but has already established overseas production capacity, providing some buffer.
In the short term, the impact of this news stems more from risk appetite rather than an immediate earnings cliff. The future holds great uncertainty, and the logic could flip at any time.
03
So, will this draft regulation actually be implemented?
Here, we need to distinguish between two scenarios.
The first scenario: The draft is largely implemented, and new optical modules are directly banned from entering the U.S. Then, 1.6T and next-generation new products would essentially lose entry qualifications for the U.S. market; for existing, certified 800G products, it would depend on whether the rules set a grace period.
In this case, customers might rush to stock up, accelerating shipments before the ban takes effect, followed by a contraction in new product orders.
However, other global regions, including Europe, the Middle East, Southeast Asia, plus domestic computing power construction in China, would serve as important buffers, preventing a complete loss of growth momentum.
Simultaneously, North American major customers would accelerate validation and introduction of U.S.-based manufacturers, but the capacity shortfall would constrain the actual replacement pace.
The second scenario: After lobbying and negotiations, the draft is significantly weakened, merely raising certification barriers without implementing a direct import ban.
Then it would likely become a case of "bad news is fully priced in," with sector sentiment recovering and market focus returning to the global AI computing power demand itself.
One point should not be overlooked: U.S.-based cloud companies themselves are stakeholders in this matter.
Forcibly cutting off mature Chinese supply chains comes at the cost of increased computing hardware costs and delays in AI data center expansion schedules. Therefore, these commercial giants, along with NVIDIA, will participate in the policy lobbying game. This policy is not purely unilaterally decided by the administrative side.
Furthermore, current U.S. data center construction is already facing severe delays. Nearly half of the AI data center projects planned to be operational in the U.S. by 2026 have been delayed or canceled, due to reasons including power infrastructure shortages and restrictions on supplies of Chinese-made components.
The construction cycle for U.S. data centers averages 3 years, far longer than China's construction speed.
If the ban is implemented, it would further exacerbate supply-demand tensions in the already severely bottlenecked optical module market, directly increasing construction costs and extending delivery cycles for U.S. AI data centers.
This is also something the U.S. government is unwilling to see, as it would seriously hinder the U.S.'s leading advantage in AI competition.
04, Conclusion
Looking at the bigger picture, the signaling significance of this rumor far outweighs its immediate practical impact.
In the past, U.S.-China technological rivalry focused on computing chips like GPUs; now, the battleground extends to underlying hardware for computing infrastructure—optical modules.
This means the entire AI data center industry chain has fully entered the arena of geopolitical competition.
The global optical module market, where China supplied the world in the past, is now moving towards regional supply chain fragmentation: The U.S. market pursues a domestic + ally supply chain, while domestic Chinese manufacturers are forced to accelerate exploring non-U.S. markets and boost overseas production line layouts.
For the optical module sector, a crucial pillar of past valuations was the orders driven by the North American AI computing power explosion.
This news will force the market to reassess North American business with a geopolitical risk premium.
Going forward, capital will likely prefer two types of companies:
First, those with a low proportion of overseas revenue and a higher proportion of domestic computing power business;
Second, companies with fast progress in self-developed optical chips and more comprehensive overseas capacity layouts.
In summary, the ultimate effect of this ban carries significant uncertainty. It may cause some short-term impact on leading Chinese optical module companies, but it will also significantly increase the construction costs and delay risks for U.S. AI data centers.
In the long run, the ban may accelerate the "bifurcation" of the global optical module supply chain—one centered on Chinese manufacturers, serving China and the "Belt and Road" markets; the other led by U.S. companies, serving North America and its allied markets.
This fragmented landscape will reduce global supply chain efficiency, increase costs across the entire industry chain, ultimately paid for collectively by participants in the global AI industry.
However, it's important to understand that short-term market movements are more about event-driven sentiment speculation. Real fundamental changes will gradually become apparent only after clear-cut rules are officially implemented.
