Under the FCC's Shadow, China's Optical Chips in a Life-or-Death Race

Under the shadow of the FCC, the life-and-death race of China's optical chips Source: Jinduan Published: 2026-11-01 14:48 Computing power photonic chips Generate poster In this breakout battle, endurance was what mattered before; now explosive power matters more.

When the AI computing infrastructure frenzy pushed optics into the eye of the storm, optical chips—the "heart of photoelectric conversion"—were naturally seen as the next hundred-billion-level blue ocean market. From 10G to 800G, and then to 1.6T and 3.2T, with each rate iteration, the value and market space of optical chips are pushed higher.

After years of sitting on the cold bench, China's optical chip industry is also raring to go, treating it as another core battlefield for domestic substitution after semiconductors.

Just as primary and secondary financing reached its peak, a bucket of cold water was thrown in their faces: on the one hand, the shadow of FCC restrictions on the optical interconnect industry chain will not dissipate, and the regulatory risk extending from terminal modules upstream to chips continues to ferment; on the other hand, the manufacturing barriers, capacity bottlenecks, and difficulty of ramping yields of optical chips themselves are clearly less favorable than the market's previous optimistic expectations.

The once-perfect logic: Why were optical chips regarded as a golden track?

I. The prospects for optical chips are bright

Both large-model training and inference rely heavily on frequent, large-scale data exchange among GPUs within a cluster, making the interconnect bottleneck more prominent. Therefore, no matter how technologies change in other tracks, optics will always benefit. In the second quarter of this year, this was condensed into one sentence—stand in the light, believe in the light.

From an industry logic perspective, this statement is not wrong.

A frequently cited data point is: computing power grows about 3 times every two years, memory bandwidth about 1.6 times, while interconnect bandwidth only about 1.4 times—this is the famous "I/O wall." When multiple GPUs work together, transmission becomes the bottleneck, and data movement efficiency becomes the core factor constraining the release of computing power.

With the advantages of high bandwidth, low latency, and low power consumption, optical interconnect has become the core solution to break through the I/O wall—the so-called "optics in, copper out." As the core semiconductor component that realizes the mutual conversion of optical and electrical signals, optical chips sit at the very upstream of the optical communications industry chain and directly determine the transmission rate, power consumption, integration level, and reliability of optical modules.

Figure: Schematic of the I/O wall problem Source: Soochow Securities

From the BOM breakdown, optical chips account for more than 30% of the cost of optical modules, and as rates iterate toward 800G/1.6T/3.2T, the share is expected to continue rising to 60%. The gross margin of high-end optical chips can reach 50%–90%, significantly higher than the optical module assembly segment.

The global laser chip market is also showing explosive high growth, with market size expected to grow from US$2.6 billion in 2024 to US$22.9 billion in 2030, a compound annual growth rate of 44.1%. If modulators, detectors, and other full-category optical chips are included, the overall market space is even broader.

For this reason, whether in US stocks or A-shares, the highest valuations are all in the upstream of optics.

Figure: Global optical chip market size Source: Yuanjie Technology Hong Kong IPO prospectus

II. Under the FCC shadow, the valuation impact on optical chips is the greatest

Optical chips not only benefit from the industry's inflation logic; previously, Chinese companies also had an additional layer of domestic substitution logic. The epic shortage in the global upstream of optics gave many domestic companies a historic opportunity to enter.

Under the shadow of the US FCC (foreign direct product rule), optical chips are the most exposed to potential impact, and future imagination space has suddenly been greatly compressed. This can be fully understood from the BOM breakdown and the current state of the domestic industry chain.

The core of electrical chips is DSP and some analog chips, which account for a high cost share, but domestic companies simply cannot make them, so there is basically no impact on domestic industry companies.

Next, the high cost share is optical chips, followed by optical devices. Optical chips account for more than 30% of optical module costs, and as rates iterate toward 800G/1.6T/3.2T, their share will further increase in the future.

Figure: Schematic of optical module industry value split Source: Guosheng Securities

The following is a brief analysis of the competitiveness of the domestic optical industry chain:

There is no need to say much about the assembly segment of optical modules. Chinese companies have the highest global share and the strongest competitiveness, and it is difficult to break away from the Chinese supply chain in a short period;

Next is optical devices. Although less dazzling than optical modules, their strength is not weak because many companies are suppliers to overseas Corning and Fabrinet;

Then there are optical chips, the focus of this article. The degree of localization is very low, and it is also a field in urgent need of breakthrough, though progress had previously been relatively fast;

As for electrical chips such as DSP, China is basically in a state of abandonment, and attention is also very low.

Taking the words too literally, many people will analogize optical chips to the "optical version of electrical chips," believing that as long as they follow the semiconductor path, breakthroughs can be gradually achieved.

But in reality, the manufacturing logic of optical chips is vastly different from that of electrical chips. It is a typical industry of "process-driven, yield is king, asset-heavy, long-cycle." In other words, optical chips are not a scaled-down version of electrical chips, but another track of precision manufacturing.

Divided by function, optical chips can be categorized into active optical chips and passive optical chips. Among them, high-speed active optical chips are the segment with the highest technical barriers and are also the core pain point of current domestic substitution. Because material requirements differ from electrical chips (all made on silicon), optical chips are also often classified by material system.

Figure: Classification of active optical chips

Unlike electrical chips, which pursue shrinking process dimensions, performance improvements in optical chips do not rely entirely on size reduction, but focus more on epitaxial structure design and growth. In some ways, they are closer to silicon carbide (SiC), so materials and processes are highly emphasized.

Its production covers four major stages: substrate manufacturing, epitaxial growth, wafer device manufacturing, and coupling and packaging. The overall process is not particularly long compared with electrical chips, but epitaxy and grating processes must achieve nanometer-level precision, the process window is extremely narrow, and yield improvement is extremely difficult.

This directly leads to differences in industry models: optical chip manufacturers generally adopt the IDM model, covering the entire process from design to manufacturing and packaging and testing. The IDM model means higher entry barriers—large upfront capital expenditure, the need to cover the entire production line from design to packaging and testing, and the requirement for stable mass production and yield control capabilities. So when demand was relatively limited back then, there were very few players in this industry.

Figure: Optical chip manufacturing process Source: Yunling Photonics

These technical characteristics also give the entire industry three distinct features:

First, technology intensity and capital intensity are superimposed. Investment in a complete InP material optical chip production line can reach several billion yuan, and core equipment relies on imports with long delivery cycles.

Second, economies of scale come from yield rather than output. On the same production line, if yield rises from 30% to 60%, costs can be halved.

Third, the customer certification cycle is as long as about 2 years, requiring the full process from internal testing and module manufacturer verification to end-customer certification and then mass production. Customer stickiness is extremely strong, and it is extremely difficult for new manufacturers to enter the supply chain.

II. Core products are EML and CW lasers

Currently, the core products of high-speed optical chips are EML and CW lasers, which also occupy the main future market for optical chips.

Figure: Global optical chips by technology route Source: Yuanjie Technology Hong Kong stock prospectus

EML chips can be called the Everest of high-end manufacturing. They require the monolithic integration of a DFB laser and an EAM modulator on the same indium phosphide (InP) chip. The process complexity is significantly higher than that of CW lasers, with more than twice the mask steps of standard CW, and yields for high-speed products may fall from 50% to as low as 15%.

The process complexity of 200G EML is about 2.5-3 times that of 100G, and by some accounts wafer-level yield is even only 10%-20%.

At present, global EML capacity above 100G is concentrated in four manufacturers: Lumentum, Broadcom, Coherent, and Sumitomo Electric. Lumentum alone accounts for a 50-60% share. Overseas 100G EML yield is about 75-85%, while domestic yield is only half of overseas levels, and 200G EML yield is only 40-50%. Most manufacturers are still in small-batch and verification stages.

Further breaking down EML helps explain where the difficulty of optical chips lies:

First, materials are the bottleneck. InP substrates are the key base for EML. According to estimates, global effective indium phosphide substrate capacity in 2026 will be only 750,000 wafers, with a shortfall exceeding 70%. Global high-end capacity is 70%-80% controlled by Sumitomo and AXT combined.

Even if a small amount of indium phosphide substrate is obtained for trial production, there are still equipment constraints. In particular, the key equipment MOCVD is supplied globally by only two companies, Germany's Aixtron and the United States' Veeco. It takes 18-24 months from order placement to mass production, and the complete cycle for building a new InP production line can reach 2-3 years. The high-end electron-beam lithography equipment on which grating processes depend generally has a delivery lead time of more than one year and is likewise monopolized by a small number of overseas manufacturers.

Finally, there are the black-box process constraints. Epitaxial growth imposes extremely strict requirements on composition and layer thickness control; a tiny deviation can lead to performance degradation and wafer scrapping. There is no shortcut to yield ramp-up; it can only rely on accumulating process know-how through repeated trial and error, a process that often takes years.

Because EML has the aforementioned high barriers, good competitive landscape, and outstanding profitability, current overseas leading manufacturers are focusing their limited energy more on EML chips.

Another segment growing faster is CW lasers. Because they pair well with silicon photonic chips (the 70mW/100mW CW lasers corresponding to current mainstream 800G and 1.6T silicon photonics solutions), and compared with EML, their process chain is shorter and yield is easier to optimize, this is the main direction in which domestic manufacturers are concentrating their efforts.

At present, the domestic substitution progress for CW lasers is somewhat better, and the landscape has formed tiers: 70mW low-power products have large volumes, are relatively mature, and have the fastest-growing number of players; 100mW products are still somewhat tight, and the shortage may persist until 2027; high-power products above 150mW target NPO/CPO and OCS scenarios, with significantly higher technical barriers, lower yields, and fewer qualified manufacturers.

In addition, the invisible customer certification barrier is equally solid. Optical chips are a typical ToB industry, and North American cloud manufacturers have extremely high thresholds for supplier selection. Once the supply landscape is formed, it is very difficult for new entrants to break through. This means that even if domestic manufacturers meet technical standards, it will be difficult for them to enter the mainstream North American supply chain in the short term.

Interpreting the industry implications behind US restrictive policies

Having understood the technical characteristics of optical chips, we can finally understand the deeper meaning behind the US FCC's restrictive policies from an industry logic perspective.

As a sword hanging over the head of the domestic optical chip industry chain, the impact of the restrictions is not a one-size-fits-all short-term cliff-edge shock, but a phased, sector-by-sector gradual squeeze. The ultimate expected result is: to extremely compress the window period for domestic substitution, and in this process, the manufacturing capacity of US manufacturers can be improved.

I. First, buy time for US companies to expand production

In the short term, the substantive impact of the policy on the industry is very limited. Leading module manufacturers such as Zhongji Innolight and Eoptolink have not been placed on the restriction list, so supply will not be affected over the next 1–2 years. Restrictions are more likely to be implemented in phases starting with the 3.2T generation, which gives the industry a 2–3 year window for structural adjustment.

This is mainly because Chinese manufacturers account for nearly 70% of global 800G/1.6T optical module shipments. The North American supply chain will find it difficult to achieve a "painless replacement" in the short term, so the restrictive measures are unlikely to be self-defeating.

The real long-term impact lies in upstream components. If implemented according to the model estimated in the Morgan Stanley report—namely, that only with "about 65% US-origin BOM value share" in the optical module system can a compliance exemption be obtained—that would mean the space for introducing Chinese optical chips and components would be greatly compressed, and the threshold for entering the already difficult North American supply chain would be raised further.

If the pace of policy is compared with the capacity expansion progress of overseas manufacturers, the intent of the policy is very clear: to buy a time window for the capacity expansion of US optical chip companies.

Currently, Coherent is upgrading its Sherman, Texas plant into the world's first large-scale 6-inch InP platform, with official guidance of a 4x surge in wafer capacity in the future, deeply tied to Nvidia's supply chain; Lumentum's newly acquired Greensboro plant is also being converted into a 6-inch InP base, expected to reach full volume in mid-2028 for mass production of high-power EML and CW light sources.

The timing of the new capacity coming online at these two giants highly coincides with the node for large-scale 3.2T volume ramp-up, and the policy restrictions happen to free up a market buffer period for them.

Source: CITIC Securities

II. China's supply chain is in a critical substitution period

For high-end optical chips, 2026 can be considered the first year of domestic production. Domestic manufacturers have previously explored a relatively clear path to breakthrough: first tackle CW light sources (combined with silicon photonics), then gradually break through 100G/200G EML.

As mentioned earlier, the CW light source process chain is shorter, mass-production yield is easier to optimize, and it has strong synergy with China's huge optical module industry. At present, domestic manufacturers have achieved large-scale mass production in the 70mW and 100mW CW laser field, and some products have entered the supply chains of leading module manufacturers, with rapid growth momentum; Yuanjie Technology's financial reports in recent quarters are a case in point.

EML is much more difficult, and the high-end market is highly concentrated. In the 2025 global laser chip market, the top five overseas companies together accounted for about 64.8% of the share, while the highest-ranked domestic manufacturer, Yuanjie Technology, had only a 3.1% market share, mainly concentrated in mid- to low-end products.

For example, Changguang Huaxin has mass production information for 100G EML, and Source Photonics under Dongshan Precision has also entered large-scale mass production for 100G, while 200G EML is only at the sample submission stage; its global optical chip market share is also just over 4%.

Source: CITIC Securities

Overall, domestic manufacturers have gained a firm foothold in the mid- to low-end. They have achieved breakthroughs in the CW field, but in high-end EML and high-power CW, there is still a 1-2 generation technology gap with overseas leaders, and shortcomings in customer certification and mass-production yield remain prominent.

What is worrying is that, due to FCC restrictions, the window period for domestic substitution is a precious and brief opportunity from 2026 to 2028.

On the other hand, overseas leading manufacturers, constrained by EML capacity, generally allocate capacity according to the priority of "200G EML→100G EML→CW lasers," giving priority to higher-margin EML products. This means that between 2026 and 2028, the CW laser market has not yet formed a monopoly pattern of strong overseas alliances, leaving a rare market gap for domestic manufacturers.

Figure: Global (2025) optical chip share for 400G and above optical interconnect products Source: Dongxing Securities

Conclusion

I. The cautionary tale of SiC

As mentioned earlier, in terms of industrial manufacturing logic, optical chips differ from electrical chips and are more similar to SiC. So any discussion of the future of optical chips cannot avoid the cautionary tale of silicon carbide (SiC).

Five years ago, SiC was still regarded as the next golden track in semiconductors, but within just two years it slid rapidly from supply shortage to full-blown overcapacity and a price collapse. In the past two years, 6-inch substrate prices fell 64.5% and 8-inch prices fell 79%; industry leader Wolfspeed entered bankruptcy reorganization due to aggressive expansion, low capacity utilization, and demand falling short of expectations.

Back then, SiC was also considered to have extreme bottlenecks in both equipment (crystal growth furnaces) and processes. But as technology spilled over and equipment bottlenecks were removed, aggressive corporate expansion led to concentrated capacity release, and scarcity suddenly turned into oversupply.

Of course, in the short term optical chips are still extremely scarce. But entering 2028, the industry will first move from comprehensive shortage to structural differentiation: low-end oversupply, high-end still tight. And EML is more prone to periodic oversupply than CW, especially 100G EML, because it faces dual pressure from generational upgrading and silicon photonics substitution. CW lasers, due to the dense influx of domestic companies, may also face the risk of mid- to low-end oversupply.

There are quite a few domestic optical chip manufacturers, but they are generally small in scale with scattered technical resources. Many localities, driven by investment attraction needs, have launched optical chip projects one after another, including many low-level redundant construction projects. If industry resources all cluster around the lower-threshold 70mW and 100mW CW and mid- to low-end EML, overcapacity and price involution will soon emerge.

Yet for the high-end EML, high-power CW, InP substrates, core equipment and other segments that truly require long-term investment, few are willing to settle down and pursue breakthroughs.

The long-term risk lies in the overseas markets repeatedly mentioned above being closed to Chinese chip companies. As a result, domestic manufacturers' technological iteration lacks validation feedback and economies of scale from major downstream customers, making chips hard to produce. Even if high-end chips are produced through various means, without mainstream customers using them in volume, it is impossible to optimize yield and reduce costs through volume ramp-up, and ultimately there is no economic value.

II. The breakout of optical chips is no longer a protracted war, but a life-or-death race

The road ahead for domestic optical chips is indeed difficult.

The first hurdle is the process barrier of high-end manufacturing. The difficulty lies in the chokeholds on core equipment and materials, the high walls of overseas customer certification, and the uncertainty of technology route switches. But for China's high-end manufacturing breakthroughs, many fields have completed the leap from lagging behind to catching up.

So process barriers are not a hard constraint for hardworking and capable Chinese. For example, the replacement window for CW lasers is right before our eyes. China has the world's most complete optical communications industry chain, from midstream modules to downstream equipment, from passive components to active chips.

But if the process barrier is combined with a 2-3 year time limit, the difficulty rises exponentially.

Because the optical module speeds required by domestic downstream customers are 1-2 generations lower than overseas, and the price gap is large. Meanwhile, the open window in the North American market is only 2-3 years, and our industry needs to go from a catching-up state to at least being not far behind overseas within such a short time—the difficulty is imaginable.

In this life-or-death race, what is also awkward for the industry is that companies with chip capabilities such as Yuanjie Technology, Huagong Technology (vertically integrated), and Accelink Technologies (vertically integrated) are still queuing for Hong Kong stock listings, only to run into such a huge shock, casting a shadow over IPO fundraising.

How to settle down to polish processes, raise yields as quickly as possible, and secure certifications are questions these companies must answer in their roadshows. In this breakout battle, endurance used to be the key; now explosive power matters more.

This article is reproduced from media reports or online platforms and represents the author's personal stance or views. Our reproduction is solely for sharing and does not mean we endorse or agree with it. If the source attribution is incorrect or your lawful rights and interests are infringed, please contact customer service promptly, and as a neutral platform service provider we will promptly correct, delete, or handle it in accordance with the law. Comments: No user comments yet. Latest News More Related Recommendations

Has the turning point for smart homes arrived? What is truly worth watching is these chips 2026-11-01 14:57

Changxin investor relations activity: discloses industry trends, price trends and technological progress 2026-11-01 14:55

STMicroelectronics Q3 profit nearly doubles, new CTO to take office next year 2026-11-01 14:52