Lumentum's Revenue Doubles: How Much Fundamental Support Backs the Frenzy for "Yi Zhong Tian"?
Over the past year or so, if one were to identify the most recognizable portfolio in the A-share AI hardware market rally, "Yi Zhong Tian" (a market nickname for Eoptolink, InnoLight, and Suzhou TFC Optical Communication (TFC)) would be almost impossible to bypass.
The so-called "Yi Zhong Tian" refers to Eoptolink, InnoLight, and TFC. These three companies are respectively positioned at critical nodes of AI optical interconnects, such as high-speed optical modules and optical components, making them one of the most capital-favored targets in this round of computing infrastructure rally. Entering 2026, this fervor has further intensified, but the market trend has clearly shifted from a one-sided surge to high-level, high-volatility trading. By the end of July, InnoLight, Eoptolink, and TFC all experienced significant pullbacks from their previous highs; among them, TFC nearly halved from its high on June 3 to July 28, while Eoptolink fell about 34% from its early July peak during the same period. On August 5, a market rumor regarding the potential U.S. restriction on importing Chinese data center optical modules triggered a collective intraday plunge for "Yi Zhong Tian," with the combined market capitalization of the three companies evaporating by approximately 128.9 billion RMB during the morning session. However, on August 11, all three stocks staged a simultaneous "deep V" recovery, with Eoptolink surging over 4% at one point during the trading session.
Such drastic volatility itself indicates that what the market is trading in "Yi Zhong Tian" today has changed compared to a year ago.
The early logic was relatively straightforward: North American cloud providers increased their AI infrastructure investments, GPU cluster expansions drove the demand for 800G high-speed optical modules, and earnings growth propelled valuation re-ratings. Now, market focus has shifted forward to 1.6T, 3.2T, CPO (Co-Packaged Optics), silicon photonics, and how high the growth rate of AI capital expenditures can be sustained over the next few years. On the morning of August 12, unverified rumors about NVIDIA potentially making strategic investments in InnoLight and Eoptolink quickly sparked market attention; both companies subsequently stated they were unaware of or preferred not to comment on the relevant rumors.
The question thus becomes more direct: After significant valuation re-ratings in stock prices, how much support can the industrial fundamentals behind "Yi Zhong Tian" actually provide?
At this very juncture, Lumentum's recently released Q4 FY2026 earnings provide an observation window from the other end of the global optical communication industry chain.
Core Financial Metrics of Lumentum in Q4 FY2026, Source: Compiled by eefocus Research Institute
In Q4 FY2026, Lumentum's revenue reached $1.0063 billion, a year-over-year (YoY) increase of 109.3% and a quarter-over-quarter (QoQ) increase of 24.5%; the non-GAAP gross margin reached 50.4%, compared to 37.8% in the same period last year; the non-GAAP operating margin improved from 15.0% in the same period last year to 36.6%. For the full year FY2026, revenue reached $3.014 billion, a YoY increase of 83.2%, of which component business revenue was $2.0056 billion, up 79.7% YoY; system business revenue was $1.0084 billion, up 90.7% YoY.
The growth has not stopped. Lumentum expects Q1 FY2027 revenue to reach $1.225 billion to $1.275 billion, with the non-GAAP operating margin further improving to 39.5%–40.5%. Calculated based on the midpoint of the guidance, next quarter's revenue will still grow by approximately 24% compared to Q4.
The author believes that observing Lumentum's earnings report alongside "Yi Zhong Tian" holds greater significance: Chinese and American optical communication companies are simultaneously validating the intensity of AI optical interconnect demand from different positions in the industry chain.
Profit growth for domestic companies has also materialized in their financial statements. InnoLight achieved an operating revenue of 19.496 billion RMB in Q1 2026, a YoY increase of 192.12%. The company explicitly stated in its quarterly report that this was mainly driven by strong investments from end customers in computing infrastructure and the continued growth in product shipments. Eoptolink expects its net profit attributable to shareholders for the first half of 2026 to reach 7 billion to 8 billion RMB, a YoY increase of 77.56%–102.93%, attributing the growth to the sustained investment in AI-related computing power and product structure optimization. TFC expects a net profit attributable to shareholders of 1.124 billion to 1.304 billion RMB for the same period, a YoY increase of 25%–45%. Its announcement also pointed out that global AI and data center construction is driving the continued growth in demand for high-speed optical components.
It can be said that, at least based on the currently disclosed financial data of Chinese and American companies, AI capital expenditures are genuinely translating into revenue and profits across the optical communication industry chain.
However, Lumentum's position in the industry chain is not entirely identical to that of "Yi Zhong Tian." InnoLight and Eoptolink are closer to the mid-stream high-speed optical module segment, while TFC covers more of the high-speed optical components and optical engine-related segments; Lumentum, on the other hand, extends from upstream lasers such as 200G EML (Electro-absorption Modulated Laser) and CW (Continuous Wave) Lasers, all the way to high-speed optical modules, OCS (Optical Circuit Switches), and CPO/NPO-related light sources and systems.
Lumentum's latest disclosed data further reinforces this judgment: Q4 shipments of 100G and 200G EMLs both set company records, with 200G EMLs already contributing over 25% of EML revenue; sales of 200G lane-speed CW Lasers continue to expand; initial shipments of 1.6T optical modules have begun; narrow-linewidth lasers grew over 130% YoY, pump lasers grew by 80%, and the OCS business is also continuing to ramp up. This means that the current high prosperity of AI optical interconnects is no longer confined to "selling a few more 800G optical modules." Demand is simultaneously diffusing upstream to high-speed light sources, 1.6T modules, and further downstream to optical switching systems. Lumentum's earnings report, showing doubled revenue, proves that this industrial cycle is currently still supported by real orders, real revenue, and real profits.
The True Bottleneck Behind 1.6T Modules is High-Speed Light Sources
Lumentum operates lasers, optical components, optical modules, and system products simultaneously. Based on the latest earnings report, as the proportion of AI-related high-value products increases, the company's overall profitability structure is undergoing significant changes. Lumentum's non-GAAP gross margin in Q4 FY2026 reached 50.4%, an increase of 12.6 percentage points from 37.8% in the same period last year; the non-GAAP operating margin also rose from 15.0% to 36.6%.
Full-Year FY2026 Revenue Structure of Lumentum, Source: Compiled by eefocus Research Institute
In its Q4 earnings call materials, Lumentum disclosed that shipments of both 100G and 200G EMLs set company records, with 200G EMLs already contributing over 25% of EML revenue; meanwhile, 200G lane-speed CW Lasers have also begun to be sold to more optical module customers. The company has also officially initiated initial shipments of 1.6T optical modules, some of which directly utilize internally supplied CW lasers.
Key Business Progress of Lumentum in AI Optical Interconnects, Source: Compiled by eefocus Research Institute
Optical module companies can expand capacity by adding automated equipment, packaging lines, and testing capabilities, but the number of manufacturers truly capable of stably and massively supplying 200G EMLs and high-performance CW Lasers remains limited. The rapid increase in the revenue proportion of 200G EMLs at Lumentum now is itself a strong industrial signal: after 1.6T begins to scale up, the value in the supply chain is shifting towards 200G/lane high-speed light sources. In the 800G era, 100G/lane had already become an important technical foundation; entering the 1.6T era, 8×200G has become one of the key implementation paths. After the total bandwidth of optical modules doubles, it is not just the DSP and packaging that need upgrading; the lasers themselves must also re-balance higher modulation rates, higher output power, power consumption, reliability, and yield. For EMLs, this involves InP (Indium Phosphide) epitaxial growth, wafer fabrication, device structure design, and long-term reliability verification. The speed of capacity expansion obviously cannot simply replicate ordinary electronic assembly production lines.
This is also why this round of AI optical communications cannot be viewed solely through the lens of "who can produce 1.6T modules."
From a business model perspective, Lumentum can simultaneously capture growth across multiple segments of the optical interconnect industry chain:
If a module manufacturer purchases Lumentum's 200G EML, Lumentum earns money from upstream optical chips; if a customer adopts a silicon photonics solution, the company can also sell CW Lasers; if a customer directly purchases 1.6T modules from Lumentum's Cloud Light, the company extends its value into mid-stream modules. In other words, under different 1.6T technology routes, Lumentum is attempting to retain its position.
This layout also holds reference value for China's optical communication industry.
Over the past few years, companies such as InnoLight and Eoptolink have entered the global first tier relying on their scale manufacturing, yield control, and customer delivery capabilities in high-speed optical modules. However, Lumentum's overall non-GAAP gross margin of over 50% reminds the market that only by mastering the scarcer upstream core chips and high-speed light sources can one control the high-value segments in the supply chain.
A More Critical Change Than 1.6T: Optics is Moving from "Modules" into Network Systems
If 1.6T solves "how fast a single link can run," then OCS and CPO/NPO represent another direction of optical technology: the changing organizational methods of AI clusters.
In Q4 FY2026, Lumentum's component business revenue was $649.4 million, a YoY increase of 102.7%; system business revenue was $356.9 million, a YoY increase of 122.6%, with the latter's growth rate surpassing that of the component business. For the full year, system business revenue reached $1.0084 billion, a YoY increase of 90.7%, also higher than the component business's growth rate of 79.7%. This does not mean that Lumentum has transformed from an optical component company into a network equipment company, but it at least indicates that the value of AI optical interconnects is extending towards the system layer.
Among these, the most worth observing is OCS, namely Optical Circuit Switches.
Traditional data center networks heavily rely on switching chips for packet forwarding, with optical modules primarily responsible for converting electrical signals into optical signals for transmission between servers, switches, and racks. As AI cluster scales expand, internal data traffic grows rapidly, and high-speed SerDes, switching ASICs, and the optoelectronic conversion itself all bring about power consumption and system complexity. OCS offers an alternative approach: directly establishing optical paths in specific scenarios, allowing data to bypass some traditional electrical switching stages.
In its Q4 earnings call materials, Lumentum explicitly stated that the ramp-up of the OCS business is still proceeding as planned, and demand is supported by a multi-year, multi-billion-dollar procurement agreement. This means that OCS is no longer just a network architecture discussion in the lab, but is beginning to enter the stage of actual procurement and scaled capacity expansion.
Of course, this does not mean that OCS will quickly replace traditional Ethernet switches.
AI training clusters require highly flexible packet scheduling, whereas OCS is more suitable for scenarios with relatively stable connection relationships, massive bandwidth demands, and a desire to reduce intermediate conversion overhead. The more likely future form is the long-term coexistence of OCS and traditional packet switching, each undertaking tasks at different network layers.
However, from the perspective of value distribution in the industry chain, this change is highly significant.
In the past, optical communication companies derived their revenue mainly from lasers, optical components, and optical modules. If OCS begins to enter AI data center networks, the market in which optical vendors can participate will further expand from "a single optical link" to "how the entire network is connected."
CPO/NPO pushes this logic one step further.
Lumentum disclosed in its earnings report that shipments of ultra-high-power CPO lasers are growing, and the company expects this business to begin making a meaningful revenue contribution by the end of the 2026 calendar year; meanwhile, Lumentum has already secured initial orders related to ELS (External Light Sources) and is participating in multiple NPO (Near-Package Optics) projects.
The judgment provided by the company's management is particularly noteworthy: these changes mean that optical technology has begun to penetrate into in-rack connectivity.
In the past, optical connections primarily addressed long-distance communication between racks, between switches, or between data centers. Inside server chassis and near GPUs and switching chips, a large number of connections still rely on copper cables and high-speed electrical signals.
However, as the I/O bandwidth of individual GPUs, ASICs, and switching chips continues to increase, the pressure on high-speed electrical connections regarding distance, power consumption, and signal integrity is also growing. Consequently, optical interconnects are beginning to move closer to computing nodes: from inter-rack optical modules to NPO, and then to CPO, light sources and optical engines are getting closer and closer to computing chips.
Short-distance electrical connections, which originally did not belong to the optical communication market, may also gradually become a new market for optical components, optical engines, and external light sources. This is also the core logic behind the "expanding optical TAM (Total Addressable Market)" emphasized by Lumentum's CEO in the latest earnings report. The company's simultaneous deployment of ultra-high-power CPO lasers, ELS, and NPO is not because it has determined which technology route will ultimately win, but rather an attempt to ensure that regardless of whether future optical connections remain in pluggable modules, move closer to switching ASICs, or further enter near-chip packaging, Lumentum can continue to provide light sources.
Finally, if the future growth of optical interconnects is merely an upgrade from 800G to 1.6T, market competition will still mainly revolve around the market share, cost, and delivery capabilities of high-speed optical modules; however, if OCS, NPO, and CPO are gradually commercialized, the competitive boundaries of the industry chain will continue to expand, and the competitors that optical module manufacturers need to face will extend from peers to laser, silicon photonics, switching ASIC, DSP, and system solution providers.
For the capital market, this also explains why the imaginative space for CPO today is far greater than that of a single generation of ordinary optical module upgrades.
The Future of "Yi Zhong Tian": Global Optical Interconnect Competition Has Begun to Redivide Labor
Collectively referring to Eoptolink, InnoLight, and TFC as "Yi Zhong Tian" can easily create an illusion: that all three companies are solely in the business of AI optical modules.
In reality, their positions in the industry chain are not entirely identical. The core competitiveness of InnoLight and Eoptolink is more concentrated in high-speed optical modules and next-generation optical interconnect products, while TFC leans more towards optical components and optical engine platforms.
Analysis of Lumentum's Major Competitors, Source: Compiled by eefocus Research Institute
If the current AI optical interconnect competition is broken down, three increasingly clear tiers can be roughly seen:
The first tier remains high-speed optical modules, which is currently the segment where Chinese manufacturers have the most significant scale advantage.
Taking InnoLight as an example, its official products already cover 1.6T OSFP224, of which the 1.6T DR8 adopts 8-lane 200G PAM4 electrical and optical interfaces, directly targeting 1.6T Ethernet, data centers, and cloud networks. Eoptolink's product roadmap, on the other hand, has continued to extend from traditional high-speed pluggable modules towards higher bandwidth and solutions closer to the system side. During OFC 2026, the company publicly showcased products such as 400G/lambda 1.6T DR4, 6.4T NPO, and 12.8T XPO, among which the 6.4T NPO utilizes 32-lane 200Gbps channels and silicon photonics technology.
From this, it can be seen that China's leading optical module manufacturers are no longer satisfied with simply repeating the product cycle of "upgrading from 800G to 1.6T," but are actively extending towards next-generation forms such as NPO and XPO.
Eoptolink has already publicly showcased its NX200 and NX300 series OCS products this year, and the company's description of their application scenarios directly points to large-scale AI clusters. It can be said that domestic high-speed optical module enterprises are also attempting to transition from "selling connection modules" to "participating in network architecture."
The second tier of competition, meanwhile, is concentrated on the increasingly valuable core components inside optical modules.
Here, Lumentum and Coherent still occupy very important positions.
In the product portfolio announced at OFC 2026, Coherent simultaneously includes 200G EMLs for 1.6T, high-power InP CW Lasers, and multiple technology routes such as silicon photonics and VCSELs; its CPO-oriented products also include 400mW high-power CW Lasers, and it is expanding its 6-inch InP manufacturing capacity. This highly overlaps with the 200G EMLs, CW Lasers, and CPO external light sources that Lumentum is currently betting on.
In the past, Lumentum's most comfortable position was acting as an upstream "pick-and-shovel" provider: no matter which module manufacturer won an order, as long as they used its EMLs or CW Lasers, Lumentum could benefit from industrial growth.
However, after acquiring Cloud Light and entering the high-speed optical module business, Lumentum simultaneously holds the dual identities of "supplier" and "competitor." On the one hand, module enterprises such as InnoLight and Eoptolink may still need to procure upstream high-speed light sources; on the other hand, they must also compete with Lumentum's own module business for cloud data center customers and 1.6T market share.
The third tier of competition extends beyond the scope of traditional optical communication companies.
Broadcom and Marvell are becoming increasingly important architecture definers.
Broadcom has currently integrated switching ASICs, silicon photonics, and CPO into the same AI infrastructure platform. Its Tomahawk 6 Davisson platform reaches a switching capacity of 102.4Tbps and combines CPO with 200G SerDes; the roadmap showcased by Broadcom at OFC 2026 also includes 400G/lane optical DSPs and comprehensive connectivity solutions for next-generation AI networks.
Marvell, on the other hand, controls another key piece of the puzzle—high-speed DSPs and electrical connections. In 2026, the company expanded its 1.6T optical DSP product portfolio, covering various solutions such as TRO, gearbox, and coherent-lite on a 3nm platform; its product portfolio also includes SerDes, drivers, TIAs, and silicon photonic optical engines.
The influence of such companies differs from that of Lumentum and Coherent. Lumentum and Coherent excel at "making the light," while Broadcom and Marvell are increasingly capable of deciding how this light should be connected, how it enters switching chips, and what architecture the entire AI network adopts. This is also the most noteworthy change in the valuation logic for "Yi Zhong Tian" in the future.
What "Yi Zhong Tian" needs to prove in the next stage is no longer just whether they can continue to be the world's largest suppliers of high-speed optical modules and optical components. Rather, it is how far Chinese manufacturers can go up and down the industry chain after global AI optical interconnect competition shifts from module competition to competition in chips, optical engines, and system architectures.
Conclusion: How Much Longer Can "Yi Zhong Tian" Keep Rising?
Finally, one cannot simply view Lumentum's doubled performance as the fundamental basis for the rise of "Yi Zhong Tian."
Because earnings reports always reflect reality later than the actual situation in the capital market.
We can look back: from 2024 to 2025, the market was mainly validating whether the demand for 800G could truly be realized; after 800G truly entered scaled deployment, the focus quickly shifted to 1.6T; and by 2026, the market is already discussing next-generation AI network architectures.
What investors currently care about is how long this high-speed growth can be sustained.
Based on Lumentum's earnings report, the bandwidth upgrade of AI networks is still genuinely occurring, and optical value is continuing to diffuse from modules upstream to components and the system layer. The prosperity in this segment is still rising.
On the other hand, can China's domestic optical communication companies take the initiative amid future technological route changes, and will their industrial value be weakened?
Lumentum's development direction has already provided the answer. If they cannot keep up with this change, not only will they be unable to retain their current valuations, but they may also fall behind in future industrial upgrades.