Produced | MiaoTou APP
Author | Zhang Bo Duan Mingzhu
Editor | Ding Ping
Header image | Visual China
As long as there is stock trading, no one knows about “Yi Zhongtian”.
The demand for AI computing power has exploded, and optical module companies have become “pipeline suppliers” in the computing power infrastructure. The three leading companies, Xinyi Sheng, Zhongji InnoLight, and Tianfu Communications, have increased by approximately 11 times, 21 times, and 9 times respectively from the low point to the highest point.
Even though there are still challenges in AI storytelling,Optical modules seem to be in a rare boom resonance: Rubin platform is increasing in volume, large model clusters are moving from 10,000 cards to 100,000 cards, 1.6T is moving from certification to batch delivery, and CPO is moving from technology demonstration to engineering deployment.
But economic resonance does not mean equal rain and dew. Although cluster expansion increases the demand for optical connections, CPO may reduce the number of traditional pluggable modules and transfer some of the value to silicon photonic engines, external laser sources, fiber arrays, and advanced packaging and test equipment. The expansion of the industry’s total market share and the decline in the value share of traditional modules may occur at the same time.
For example, the total water consumption in AI cities is soaring, but the business of “selling plug-in pipes” may shrink, and profits are flowing to water pumps, valves, joints and testing equipment.
The redistribution of profits means that seats may be rearranged. For second-tier optical module companies, in addition to benefiting from the industry boom, this is also an opportunity to strive for a new position.
The most real opportunity at the moment comes from the supply and demand gap of 1.6T; superimposed on the FCC (Federal Communications Commission) After the implementation, cloud vendors proactively introduced second and third suppliers.Second-tier optical module manufacturers have obtained a customer certification window that was previously extremely difficult to obtain.。
Funds are already moving. The valuation of first-tier leaders has been pushed to about 100 times. After being over-priced, the short-term cost performance is obviously not high, while the financing activity of some second-tier targets is increasing. On the opening day of the Optical Expo on September 9, Huagong Technology hit the daily limit within seconds of opening, and Cambridge Technology rose by more than 16% in half a month. The second-line elasticity was significantly greater than the first-line.
Opportunity comes, but the window won’t stay open.
Opportunities for second-tier optical module manufacturers will ultimately depend on which companies can convert increased demand into orders, profits and improved customer status when the optical interconnection value chain is re-segmented. This article aims to answer exactly this question.
How much room is there for upwards?
The certainty of the future growth space of the optical module industry isIt is the highest level among all current AI hardware segments.。
According to LightCounting data, the CAGR of the global optical module market from 2025 to 2030 (CAGR) about 22%, of which the CAGR of Ethernet optical modules driven by AI computing power clusters reaches 24%; by 2031, the Ethernet optical module market will approach US$60 billion, of which sales in the AI field will exceed US$45 billion, accounting for approximately 75%.
Goldman Sachs is even more radical. The latest report estimates the global optical module market TAM from 2026 to 2028 (total addressable market) were raised to US$67.7 billion, US$131.4 billion, and US$148.5 billion, respectively 33%, 81%, and 115% higher than the previous forecast.
No matter which caliber is used,The rapid expansion of the industry cake is a deterministic event。
Structurally, high-speed products are the absolute core of growth. Goldman Sachs estimates that shipments of 800G and above optical modules will reach 78.22 million, 144 million, and 171 million respectively from 2026 to 2028; of which 1.6T will increase from 32.83 million in 2026 to 71.36 million in 2027 (Note: 800G, 1.6T, etc. are classified according to the single port transmission rate of the optical module. The number represents the amount of data that can be transmitted per second. The higher the rate, the more data is transmitted per unit time, and the higher the technical threshold and unit price are.）。
From a monetary perspective, 800G and above optical modules will account for approximately 67% of the overall optical module market in 2026.
However, the responsiveness of the supply side is far from keeping up with the expansion of demand.
The core constraints come from upstream materials: the overall EML gap in the entire industry is about 25%-30%, of which the overall industry gap for 200G high-end EML laser chips in 2026 is close to 70%; DSP chip delivery time has been extended to about 50 weeks, and most orders are long-term, legally binding, and require prepayment.
Affected by this, even front-line leaders such as “Yi Zhongtian” have significantly restricted their Q1 delivery pace.
不過，The larger the gap between supply and demand, the more opportunities there will be for second-tier optical module manufacturers.。
According to industry research, the current gap between 800G supply and demand is small, and the supply pattern has been solidified. Zhongji InnoLight’s market share in 800G products even exceeded 40% at its peak. Second-tier manufacturers in this category have neither material advantages nor time windows for customer certification.
The real opportunity for second-tier manufacturers comes from the 1.6T gap space. The 1.6T material bottleneck is hard, and the production capacity of first-tier manufacturers is locked by 200G EML and 3nm DSP, unable to meet all demand, and the gap is passively spilled over to the second-tier manufacturers.
According to agency estimates, the actual global demand for 1.6T optical modules in 2026 is about 25-30 million units; however, the caliber of shipments deviates greatly, with conservative estimates of only 11-22 million units, and the most aggressive Goldman Sachs forecast of 32.83 million units.
If a conservative caliber is used, the gap for 1.6T is about 3-19 million; if the mid-range is conservative, the gap is about 8-19 million.
Acceptance rate low gap (8 million pieces) median (13.5 million pieces) high gap (19 million pieces) 20% 16027038030% (baseline) 24040557040% 320540760
Note: Estimated based on public data
Based on the baseline scenario, the available space for second-tier manufacturers is about 2.4-5.7 million units per year. This means that second-tier manufacturers have entered a 100-billion-dollar market from scratch.
Taking Cambridge Technology as an example, if it is allocated 1.5 million (Approximately 9.8 billion yuan in revenue), equivalent to its 2025 revenue (48億元), which is the order of magnitude of recreating Cambridge Technology.
However, it should be noted that for second-tier manufacturers, thisIt’s just a phased capacity bonus, once supply recovers, orders may still return to first-tier manufacturers with higher yields and more stable delivery.
Therefore, only if second-tier manufacturers use this round of capacity expansion to enter the head customer system and form irreplaceable capabilities in silicon photonics, external light sources, FAU or testing links, will they be able to achieve sustained growth in share.
The real verification criterion is that four conditions are true at the same time:
-
客戶認證: Whether it is possible to proceed from sample delivery to batch delivery;
-
量產良率: Whether it can be stabilized at an industry acceptable level;
-
Core material lock: Can production capacity be locked in?
-
operating cash flow: Whether it can keep up with the profit growth rate, customer certification progress, overseas production capacity implementation pace and material support capabilities.
Therefore, the prosperity of the industry is just the background, and who can take away how much profit needs to be analyzed in detail.
Who is more qualified?
The most critical gap between the first-tier and second-tier optical module manufacturers lies in whether they are deeply bound to North American cloud vendors such as Google, Microsoft, and Amazon. Zhongji InnoLight has gradually become the big brother of domestic optical modules by virtue of its deep binding with Google and Microsoft starting from the 400G era.
Nowadays, the fragmented demand in the supply chain brought about by FCC policies and the gap between supply and demand of optical modules are a historic window for second-tier manufacturers to obtain “tickets of admission.”
Among second-tier optical module manufacturers, there are 5 worthy of study:Huagong Technology, Cambridge Technology, Changxin Bochuang, Dekeli, Huashengchang。
But these five companies actually go their own separate ways. Huagong Technology and Cambridge Technology compete for customers and production capacity share of high-speed modules, but the former competes for platforms and the latter competes for profit elasticity; Changxin Bochuang and Dekeli avoid direct competition, one is a card-position interconnect device, and the other is betting on DCI and OCS; Huashengchang simply does not focus on the main track of optical modules, but sells test equipment.
They face different customers, bear different risks, and have inconsistent profit realization cycles.
Note: From public data and information
Let’s break it down house by house.
The advantage of Huagong Technology lies in its more complete coverage of the industrial chain., is its diverse customer structure (Domestic Huawei ecosystem + overseas AI incremental + traditional business) strategic choice.
It has a layout from chips and devices to modules and systems; yes華爲It has been a gold medal supplier for 15 consecutive years, providing 800G/1.6T/3.2T and other high-speed optical modules and optical engines in the Ascend AI cluster. It is also the first A-share company to release and implement the world’s first 3.2T NPO product. It has now entered the order delivery stage for leading customers; it also carries domestic telecommunications business, laser equipment and other diversified sectors.
Under this structure, the gross profit margin of its optical interconnection business is only 14.47%, which is about one-third of that of the leading pure module factory. It is mainly diluted by low-margin traditional businesses.
It can be said that Huagong TechnologySacrificing short-term profit elasticity in exchange for higher certainty and technical depth。
However, the quality of its profits needs to be improved. In the first half of 2026, Huagong Technology’s revenue was 7.822 billion yuan, an increase of only 2.53% year-on-year; net profit attributable to the parent company was 1.186 billion yuan, an increase of 30.17%; non-net profit after deducting 747 million yuan, an increase of only 2.42%; the difference of approximately 439 million yuan came from changes in fair value and investment income.
In other words, a considerable part of Huagong’s high profit growth is not contributed by its main business.
For Huagong Technology, the follow-up is not to see how many high-speed products can be launched, but whether self-developed silicon photonics can substantially reduce costs and whether inventory can be smoothly sold out.
Cambridge Technology is the second-tier company with the strongest profit elasticity. This also stems from its highly concentrated overseas income structure. In 2025, the company’s overseas revenue will account for 93.88%, and the top five customers will account for 86.1%. Customer coverage思科and other top global ICT equipment manufacturers. This structure determines that its fate is deeply tied to North American AI computing power capital expenditures.
In the first half of 2026, Cambridge Technology’s revenue was 2.705 billion yuan (+32.92%), 328 million yuan (+171.08%), with a deduction of RMB 319 million (+168.42%). Kuefei and Guimu are almost synchronized, indicating thatProfits come from the main business, not floating profits on the books。
Moreover, Cambridge Technology’s high-speed optical module revenue was 1.675 billion yuan, +240.85% year-on-year, with a gross profit margin of 34.49% (Year-on-year +10.27pct), the comprehensive gross profit margin rose to 30.77%. On the product side, its 1.6T product has completed certification testing, and its annual production capacity of 6 million units has been achieved ahead of schedule. It is expected to reach tens of millions within the year.
But behind high elasticity is high capital occupation. Although it made money in the first half of this year, its operating cash flow was a net outflow of 1.113 billion yuan; exchange losses were about 202 million yuan, equivalent to more than 60% of the net profit attributable to the parent company; inventories, accounts receivable, and prepayments were all increasing, and almost all of the half-year revenue was settled on the receivable end.
這說明，Overseas major customers occupy a strong position in the industrial chain, and the pressure on payment deadlines and stocking is mainly borne by Cambridge Technology。
The key to whether Cambridge Technology can enter the “quasi first line” lies in whether the yield rate can be stabilized after mass production, whether the customer structure can be improved, and whether operating cash flow can keep up with profit growth.
長芯博創It is the only company among the five companies with “profit and cash synchronized”. Through its subsidiary Changxinsheng, it cooperates with North American giants谷歌It has established a long-term and stable supply relationship and is one of the two largest direct suppliers of Google MPO in the world, with a procurement share of approximately 30%.
Changxin Bochuang is expanding its active products to Google (From passive devices that only consume no power, such as fiber optic connection, distribution, and coupling, to active products such as optical modules and active optical cables that require photoelectric conversion, power supply, and signal processing, thus entering higher value links.), the potential is huge but it does not mean that large-scale orders have been formed.
Its overseas revenue in the first half of 2026 was 1.170 billion yuan, accounting for 69.30%, and the gross profit margin was as high as 62.80%; the overall revenue was 1.688 billion yuan, an increase of 40.72%; the net profit attributable to the parent company was 321 million yuan, an increase of 91.08%; after deducting non-profit items, it increased by 94.49%. Operating cash flow +311 million yuan,Financial health。
Whether Changxin Bochuang can complete the transition depends on whether AOC/optical modules can move from small batches to millions of levels, rather than relying solely on the expansion of large customers to continue to expand the original business; whether the 4.5 billion yuan long-term agreement can be actually implemented, and whether Changxinsheng’s remaining equity injection can bring about consolidated financial statements.
德科立It is more like a “technology option”. In the past, it mainly built long-distance information highways for operators such as China Mobile and Telecom. Now it has transformed into DCI (Data center interconnection), forward business also includes OCS (optical circuit switching）。
Its revenue in the first half of 2026 was 545 million yuan, an increase of 25.78%; net profit attributable to the parent company was 98 million yuan, a surge of 249.74%; but net profit was only 22 million yuan, an increase of 16.17%. The difference of 76.05 million yuan came from non-recurring gains and losses, of which changes in the fair value of financial assets and disposal gains and losses amounted to 85.27 million yuan. Q2 single-quarter deductions amounted to RMB 8.0657 million, a year-on-year and month-on-month decrease. This showsIts profit surge does not come from its main business。
It has real layouts in DCI and OCS. The former has generated actual revenue, but its size is still small. The latter company has made it clear in investor communications that OCS is still in the prototype stage and has not been included in expectations in its financial budgets and performance plans in the past two to three years. There will be no revenue planning in 2026 and 2027.
The hard constraints are also obvious. The investment project has been postponed four times, and the non-profit deduction is weak, but the valuation is high. It also needs to prove that the Thai factory is put into production as scheduled, sample orders turn into scale revenue, and the non-profit deduction is synchronized with revenue.
把華盛昌It’s not accurate to put it in the same scoring framework as the other four. Its core business is instrumentation, and its entry into the optical communications industry chain will mainly rely on the acquisition of Galant in 2026. itWhat we share is the demand for test equipment brought about by industry expansion.。
But the logic of “selling shovels” is indeed attractive. Regardless of whether it is pluggable, NPO or CPO, as long as the speed is increased and the production capacity is expanded, the complexity of the test link and equipment investment will usually increase. Galante’s consolidated monthly revenue was 97.89 million yuan and net profit was 27.06 million yuan. Orders in the first half of the year were approximately 600 million yuan, and Q2 orders doubled from the previous month.
The equipment link is not inherently safer than the module link. Galant’s single-month profits cannot be extrapolated to the whole year; the 843.18% assessed value-added rate has formed 344 million yuan in goodwill, accounting for approximately 29% of the net assets attributable to the parent company; the three-year cumulative performance commitment is only 115 million yuan, and a net profit of 76.3 million yuan has been achieved in the first half of the year. The commitment is more like binding a team than a bottom-line profit. After excluding the consolidation, Huashengchang’s original main business income only increased by 9.34%.
因此，Huashengchang should be regarded as an equipment company in the optical communication production expansion chain., the follow-up key depends on customer concentration, core component sources, order continuity, and whether high-end equipment can truly enter the 1.6T and silicon photonics mass production lines.
Therefore, the second-tier breakthrough will not occur in the form of simultaneous advancement of the entire sector.
If we look at the 1.6T business:
Huagong Technology: The most complete layout, 800G/1.6T has been delivered in batches, 1.6T uses self-developed silicon photonic chips, and the monthly production capacity plan is increased from 40,000 to 200,000 by the end of the year. However, due to the shortage of DSP and other materials in the first half of the year, the actual delivery was 50% lower than expected.
Cambridge Technology: closest to bulk shipments. The 1.6T product has been tested at three customers and is planned to be shipped in large quantities in the second half of 2026. The annual production capacity is planned to be no less than 6 million units.
長芯博創: Instead of taking the main battlefield of pluggable modules, the 1.6T AEC is launched, equipped with Marvell 1.6T AEC DSP, for internal interconnection of AI clusters. The space depends on the cloud vendor’s choice of short-distance interconnection solution, and it is a “small but precise” differentiated card space.
德科立: The two lines have different rhythms. Datacom’s 1.6T optical module is expected to be ordered in small batches in Q4 2026; DCI direction 800G boards have been delivered in small batches, and 1.6T boards are in pre-development. Looking at DCI in the short term and data communication modules in the long term, the space is “differentiated survival” rather than a head-on breakthrough.
華盛昌: It does not produce modules, but provides 100G-1.6T full-rate test equipment through Galant.
However, a real breakthrough needs to be reflected at least in financial results: revenue growth should be accompanied by stable or improved gross profit margins, profit growth should be gradually converted into operating cash flow, production expansion should be based on verified orders rather than forward expectations, and new products should move from sample delivery and certification to batch delivery.
Without these changes, the so-called technology stuck position is still just a valuation story.
總結
The duration and conduction path of the “supply exceeds demand” state in the optical module industry are highly uncertain. Fortunately, for second-tier optical module manufacturers, the window in the next 12 to 18 months does exist, but the window does not equal victory.
Second-tier manufacturers need to accomplish four things during this period:Obtain certification, lock materials, stabilize mass production and cash recovery。
Four things are indispensable. If you only accept certification without locking materials, mass production will not be achieved; if you only accept mass production without cash, the greater the scale, the greater the financial pressure; if you only accept cash without building technical card slots, you will be left behind in the next round of generational switching; andOnce the leader completes its capacity ramp-up, the survival space for latecomers will be further compressed.。
Specifically, Huagong Technology needs to prove that the entire industry chain can improve profit margins, Cambridge Technology needs to prove that high growth can generate cash, Changxin Bochuang needs to prove that passive advantages can be extended to new value links, Dekeli needs to prove that technology reserves can be scaled up, and Huashengchang needs to prove that mergers and acquisitions bring sustainable capabilities rather than one-time profits.
