Shareholders of Zhongji InnoLight have had a rough four months. From July 1 to today, the stock price has fallen by 57.84%. Eoptolink’s situation is even worse, with a decline of 65.11%. In the first three quarters, the two companies generated revenues of CNY 69.5 billion and CNY 31.8 billion respectively, with growth rates of 178% and 93%. While their businesses are performing well, their stock prices have been halved.
On November 11, Zhongji InnoLight once again topped the list in terms of trading volume across the entire market. With the market recovering, what should investors focus on next? Is it orders for 2027 or products for 2028?
In the optical module industry, these two years represent distinct scenarios.
01、1.6T Shipments Double, but Prices Drop by 20%
The most unique aspect of the optical module business is that unit prices consistently trend downward. Selling more does not necessarily mean earning more.
To explain its function briefly: Optical modules convert electrical signals into optical signals and vice versa. They facilitate data transmission between servers, racks, and within data centers. As data centers expand, the number of connection points increases, which has been the primary driver behind the industry's prominence over the past two years.
Goldman Sachs raised its shipment forecasts for high-speed optical modules on September 8. For 800G modules, shipments are projected at 45.39 million, 49.48 million, and 48.66 million units over the next three years, with little growth expected after 2027. In contrast, 1.6T modules show steeper growth:32.83 million units in 2026, rising to 71.36 million in 2027—more than doubling. 3.2T modules begin scaling up from 2027 with 23.04 million units, reaching 67.73 million in 2028, surpassing 1.6T volumes.
Combining 1.6T and 3.2T shipments, total volumes reach 94.4 million units in 2027 and 122.58 million units in 2028, representing a 30% increase. Demand has not disappeared; it has merely shifted from 800G to 1.6T, and then to 3.2T.
According to the same forecast set, the average selling price (ASP) of 800G modules drops from USD 418 to USD 314, while 1.6T ASPs fall from USD 798 to USD 600—a quarter decrease over two years for both categories. Although shipment volumes rise by 30%, unit prices drop by a quarter. When netted out, revenue growth will be significantly lower than shipment growth.
This price erosion is not new. It occurred when 400G replaced 100G, and again when 800G replaced 400G. New generations launch at peak prices, but as volumes ramp up and competitors enter, prices decline. Optical module manufacturers capture profits primarily during the first two years of a technology transition; thereafter, success depends on who can reduce costs faster.
This places immense pressure on cost management. Zhongji InnoLight reported a net profit attributable to parent company shareholders of CNY 22.238 billion for the first three quarters, with a net margin of 31.98%. Eoptolink reported CNY 11.372 billion in net profit, with a net margin of 35.71%. These margins are robust for manufacturing, but thicker margins are less resilient to price cuts. If cost reductions lag behind price declines in any given year, the impact will be clearly visible on the income statement.
There are two strategies to withstand price declines. First, move up the product ladder: 3.2T modules command higher prices than 1.6T, which are higher than 800G. Shifting the product mix upward helps support average prices. Second, adopt alternative solutions to cut costs. Approximately 60% of 800G modules use silicon photonics. Institutional estimates suggest that the share of silicon photonics in 1.6T modules will rise to 74% by 2028. Silicon photonics integrates optical engines onto wafers rather than using discrete components; as volumes scale, unit costs decrease.
These figures are institutional forecasts, not confirmed orders. The true indicators to watch are the single-quarter gross margin in Q4 2026 and tender volumes in Q1 2027. If the share of 1.6T modules rises but gross margins continue to fall, it indicates that price concessions are outpacing cost reductions, meaning higher volumes cannot translate into profits.
02、95% of Revenue Comes from Overseas; Domestic CNY 100 Billion Capex Not Yet Orders
Revenue structure further illustrates the situation. In the first three quarters, Zhongji InnoLight’s revenue was CNY 69.547 billion, up 178.13% year-on-year; Eoptolink’s was CNY 31.850 billion, up 92.97% year-on-year. Regional breakdowns are not disclosed in quarterly reports, but based on interim report standards, overseas revenue accounted for 94.82% for Zhongji InnoLight and 97.92% for Eoptolink. Describing these two companies as predominantly overseas-driven is accurate.
New capital expenditure is indeed entering the domestic market. China Mobile’s 2027 investment plan includes total procurement exceeding CNY 500 billion, with computing power investment alone surpassing CNY 100 billion. In 2026, capex for computing power networks is projected at CNY 37.8 billion, a 62.4% year-on-year increase. Chairman Chen Zhongyue stated on November 6 that computing power represents the largest incremental investment for the company. China Unicom’s computing power revenue grew by 12.2% in the first three quarters, intelligent computing center revenue rose by 10.4%, and standard rack count reached 1.2 million.
However, no one can currently clarify how much of this spending will translate into revenue for Zhongji InnoLight and Eoptolink. Operators’ computing power investments primarily target servers, switches, facilities, and power infrastructure, with optical modules serving as supporting components. Procurement follows centralized bidding, where market share depends on winning bids. There is no public disclosure regarding how much share these two companies secure in domestic operator centralized procurement, nor is there public data on how much of the CNY 100 billion corresponds to optical modules. At present, this can only be considered potential demand, not confirmed orders.
The domestic market is worth monitoring because it serves as a buffer. With 95% of revenue coming from overseas, if overseas growth slows, there is nowhere else on the balance sheet to compensate. If the domestic market truly scales up, it provides at least one additional revenue source.
There is also a timing lag. Operators issue annual investment plans at the start of the year, with centralized bidding typically landing in Q2 and Q3. Optical module manufacturers recognize revenue even later. Therefore, the CNY 100 billion planned for 2027 will likely reflect in financial statements in the second half of 2027 or even 2028. If overseas capex turns downward in2027, whether domestic spending can fill the gap remains uncertain due to mismatched rhythms.
Overseas markets remain the current lifeline. The combined capex of the four major North American cloud providers in Q3 2026 was approximately USD 247 billion. Microsoft’s full-year projection is USD 170 billion, Amazon’s is USD 220 billion, and Google’s exceeds USD 200 billion, with plans to break USD 300 billion in 2027. Over the past few years, the doubling of revenue for these two Chinese companies relied heavily on these accounts. Consequently, capex guidance in the quarterly earnings reports of these four firms influences stock prices more directly than any domestic news.
Another variable emerges in 2028, specifically regarding product form factors. Industry research outlines a long-term roadmap: pluggable modules account for approximately 60%, NPO (Near-Packaged Optics) for about 10%, and CPO (Co-Packaged Optics) for roughly 30%. In scale-out scenarios, CPO penetration could reach 40% to 50%. If this trajectory holds,30% of demand in 2028 will shift away from pluggables, requiring a recalculation of capacity currently configured for pluggables. The distinction between CPO and pluggables lies in packaging location. Pluggables are independent modules inserted into switch panels, easily replaced upon failure.
CPO co-packages the optical engine directly with the switching chip, eliminating the module housing and reducing power consumption. The benefits include energy efficiency and higher density, but the trade-offs are maintenance complexity and lower yield rates. CPO replaces the module form factor, not the optical connectivity itself.
Demand estimates also vary significantly between sources. Goldman Sachs projects 1.6T shipments of 54.85 million units in 2028, whereas industry research suggests a range of 120 to 130 million units—a difference of more than double. One method extrapolates from cloud provider capex, while the other back-calculates from chip platforms and cluster ratios. A third perspective appeared in an institutional monthly report on November 10, suggesting that overseas computing supply chains face downward revisions in investment growth and increased supply, while domestic computing chains remain in an upswing. Verification will depend on large customer orders placed towards year-end. With these three viewpoints presented together, there is no consensus yet for 2028.
Why discuss 2028 now? Because capacity expansion is calculated based on demand two years out. Building production lines, ordering equipment, and preparing materials require lead times of one to two years. If the demand form factor in 2028 differs from current expectations, today’s investments may be misdirected.
Returning to the core question: Demand is still growing, but who captures the value?
Price erosion claims a portion, dropping by a quarter over two years. Form factor shifts may claim another, with CPO potentially accounting for 30% by 2028 according to research. Customers determine the third portion: With 95% of revenue from overseas for the two leaders, how much of the domestic CNY 100 billion capex translates into actual orders depends on centralized bidding results.
Demand has not disappeared; this is the most certain fact currently. The revenue growth rates and net margins of both companies remain at the forefront of the industry. What remains uncertain is whether price cuts will outpace cost reductions, whether CPO penetration will truly reach 30%, and how much share they will secure in domestic centralized procurement.
Three metrics are trackable: Single-quarter gross margin in Q4 2026, cloud provider tender volumes in Q1 2027, and winning bid shares and prices in China Mobile’s 2027 optical module centralized procurement. Only when these three numbers emerge will judgments have a solid basis.