Earnings Call Highlights: Innolight (300308.SZ) | 2026 H1 — A Full Year's Results in Six Months, but the Real Signal Is HK$53.4 Billion in Expansion Firepower

2 days ago

11 min read

Innolight posted 2026 first-half revenue of RMB 41.778 billion and net profit attributable to shareholders of RMB 13.651 billion, both higher than its full-year 2025 figures (RMB 38.240 billion / RMB 10.797 billion). Gross margin rose 6.63 percentage points YoY to 46.59%. But the numbers for the half are not the main story. Three things further down matter more: 2027 orders for 1.6T and 800G are already in hand, 2.4T/NPO/XPO are slated for volume production in 2H27, and the HK$53.41 billion raised in July's Hong Kong listing is being poured into capacity and materials. This is a company betting "this year's profits" entirely on "next year's capacity."

This article is for technology and industry trend analysis only and does not constitute investment advice.

1. Three Core Signals

The call was held on Sunday, August 23 at 8 p.m., with investor relations activity record No. 2026-008. The company was represented by Vice President and CFO 王曉麗 and Vice President and Board Secretary 王軍, and more than 150 institutions took part, with funds, brokers, foreign investors and insurance asset managers nearly all present. Holding a call on a Sunday night says a lot about how tense the market is.

Three signals:

Signal one: profit is growing faster than revenue, and consistently so. First-half revenue rose 182.49% YoY and net profit attributable to shareholders rose 241.70%. Derived Q2 revenue grew only about 14.3% QoQ, while net profit grew 38.0% QoQ. Management was direct about why: higher 1.6T shipments QoQ lifted gross margin.

Signal two: visibility has gone from "three months" to "more than a year." Management was clear in Q14: optical module makers used to sign three-month orders, but many customers have now placed orders for 2027, and some are negotiating long-term agreements. For the first time in a decade, an optical module maker can plan production by the year.

Signal three: the new-product timeline is locked into 2H27. 2.4T, NPO (near-packaged optics) and XPO are all in custom development and sampling. Management said they are "expected to reach volume production in 2H27, with larger-scale shipments expected in 2028," and that the company "is expected to be among the first suppliers to deliver NPO."

Order visibility stretching from three months to a full year is not just a sign of a good cycle; it means customers are already locking in optical interconnect as a capacity bottleneck.

2. Revenue and Financials

First, the hard data. All figures below come from the company's 2026 interim report (China Accounting Standards for Business Enterprises, RMB). Q2 figures are derived as "first half minus Q1":

Metric	2026 H1	YoY
Revenue	RMB 41.778B	+182.49%
Net profit attributable to shareholders	RMB 13.651B	+241.70%
Net profit attributable to shareholders, excl. non-recurring items	RMB 13.092B	+229.32%
Gross margin	46.59%	+6.63pp
Net margin	32.68%	+5.66pp
Basic EPS	RMB 12.31	+238.19%
Weighted average ROE	37.62%	+18.78pp
Total assets	RMB 68.942B	+52.23%

By quarter: Q1 revenue was RMB 19.496 billion and net profit attributable to shareholders RMB 5.735 billion; derived Q2 revenue was RMB 22.282 billion (+14.3% QoQ) and net profit RMB 7.916 billion (+38.0% QoQ). The gap between revenue growth and profit growth is direct evidence of gross margin improvement.

The revenue mix deserves even more attention:

• Optical transceiver modules: RMB 41.355 billion (+186.56%), 99.0% of total revenue. This is now a pure-play optical module company

• Overseas revenue: RMB 39.615 billion (+209.9%), 94.8% of the total; domestic revenue was only RMB 2.163 billion

• Other businesses: RMB 423 million (+18.3%)

With nearly 95% of revenue from overseas, the company's revenue curve is essentially a mirror of North American cloud providers' capex.

On dividends, the company will pay a cash dividend of RMB 12 (tax inclusive) per 10 shares on a base of 1.170 billion shares, for an interim payout of about RMB 1.4 billion.

3. Why Gross Margin Reached 46.59%: 1.6T Ramp and Rising Silicon Photonics Penetration

Gross margin on optical transceiver modules was 39.96% a year earlier and 46.59% this year. Management attributed the gap of more than six points to two things: QoQ growth in 1.6T shipments and rising silicon photonics (SiPh) penetration.

These are really two sides of the same thing. The biggest difference between the 1.6T generation and 800G is not just the doubled data rate but a reallocation of the BOM: the cost shares of optical engines, DSPs and packaging have all been rewritten, and whoever raises the SiPh ratio and pushes up yields captures that spread. We fully unpacked the cost structure of this generational jump in 800G vs 1.6T Optical Modules: A Generational Jump, a BOM Teardown, and Who Really Captures the Margin.

Asked whether gross margin can rise further (Q1), management's wording is worth reading word for word:

• Positives: "the company has a relatively high 1.6T market share" and "silicon photonics penetration is also expected to rise further"

• Negatives: "some upstream materials face price-increase pressure, part of which will pass through to the company," but "the company will also negotiate with customers to pass on part of the cost pressure"

• Goal: "to keep gross margin at a relatively high level and hopefully achieve a slight increase; a large increase would not be consistent with the industry's characteristics."

That last line is the most honest of the entire call. It is framed as "the internal goal is modest growth," not an empty "we will keep improving." A "goal" carries less certainty than an "expectation," and an "expectation" less than a "confirmation"; on gross margin, management chose the most conservative tier.

Management is betting the real gross margin jump on 2027: 2.4T, NPO, XPO and other "high-bandwidth, customized" products whose "gross margins are expected to be better than existing products." In other words, today's 46.59% is supported by BOM cost and yield improvements; the next leg depends on a product mix upgrade.

4. Management Outlook: 2027 Orders in Hand, 2.4T/NPO/XPO Slated for 2H27

The most valuable part of the call was not the financials but the technology roadmap answers in Q3, Q12 and Q19.

NPO positioning (Q3, Q12). Management said NPO "is expected to become an optical interconnect solution widely accepted by industry customers, mainly for scale-up scenarios," with advantages of being "easy to maintain, low cost, and highly stable and reliable," and that it "can fully leverage the mature process advantages of optical modules." That last clause is key: for pluggable module makers, NPO is an extension, not a disruption. Timeline: some customers have given clear demand guidance and orders, shipments are expected in 2027, and more customers will follow in 2028. The company describes itself as "in a leading position in development" and "expected to be among the first suppliers to deliver NPO in 2027," stressing that NPO is highly customized and "customers won't open up many opportunities," giving early entrants a structural advantage.

Division of labor between CPO and NPO (Q19). Management's view: customers have already announced volume production of CPO, mainly for scale-out, with gradual adoption in scale-up; on the scale-up side, NPO will also be a widely accepted solution; meanwhile, pluggables will continue to appear in both scale-up and scale-out, including 2.4T, 3.2T and XPO. The conclusion: "future optical connectivity will not be a single solution but a diversified set of solutions."

This view closely matches the conclusion of this year's three-way CPO/NPO/XPO panel at OCP APAC: it is not a battle over which path replaces which, but each finding its place in different scenarios. For background, see 2026 OCP APAC Summit | The CPO/NPO/XPO Panel: Not a Battle of Roadmaps, but the 409.6T Wall 18 Months Out and Is NVIDIA's Shift from CPO to NPO Bad News for the Laser Supply Chain? Four Recent Debates Explained.

Technology reserves (Q16, Q18). Management named three technologies that will go into new products: silicon photonics, coherent lite and thin-film lithium niobate (TFLN), plus hybrid integration of electronic and photonic chips. In their words, some of this technology has been accumulated "for more than 10 years," and "the technology curve is expected to become steeper and steeper."

Capacity location (Q20). Capacity for new products "will be placed more overseas," with coherent lite and NPO named specifically. In today's geopolitical environment, that sentence carries a lot of information.

Note: at the end of the minutes, the company explicitly states that "the above meeting minutes do not represent the company's profit forecast or performance guidance." The call gave no revenue guidance figures at all. All forward-looking statements were qualitative. That is normal for China A-share calls, but it means investors have to build their own models.

5. Supply Chain Clues: The Bottlenecks Are 3nm DSPs, Optical Chips and PCBs

The most genuine industry signals of the half were hidden in Q4 to Q6 and Q15.

What is short. Management named three items: optical chips, electronic chips and PCBs. The tightest is "capacity supply of 3nm DSPs." This matches what we have seen across the supply chain this year: the 1.6T bottleneck has shifted from optics to electronics, and 212.5 Gbps/lane SerDes can only be made on the most advanced process nodes.

How they are coping. The company uses a three-part toolkit: "long-term agreements, prepayments and qualifying new suppliers." The footprint in the financials is clear: inventory rose from RMB 12.681 billion to RMB 19.826 billion, with more than RMB 4 billion added in Q2 alone, and "other long-term assets" (prepayments locked in for more than a year) grew by more than RMB 1 billion QoQ in Q2. This is not inventory out of control; it is buying upstream capacity in advance.

Will price increases eat into margins? Management's answer had two layers: "some materials have indeed seen price increases, but the increases are not large for the products the company buys in large volumes," because "we have locked in stronger contracts, our order volumes are larger, and the prices we get are more favorable." It also proactively clarified one point: "the company's product price declines are relatively rational, not the sharp drops rumored in the market". This was a direct response to recent market rumors of an "optical module price war."

When will it ease? "Supply tightness is expected to persist this year, and improvement will not be easy, but it should get better quarter by quarter." Supplier capacity is "expected to be gradually released from the second half of this year to the first half of next year." A QoQ improvement is expected in Q3.

It is worth adding that the deepest point of this shortage is not actually the DSP but compound semiconductor substrates further upstream. We tracked that thread in They Can Ban Modules, but Not Substrates: The Decisive Battleground in Optical Communications Has Moved Down to Indium Phosphide.

6. Three Numbers That Look Like Bad News: Cash Flow, FX Losses and Capex

Three numbers in the interim report look alarming in isolation, but the call took them apart one by one.

Operating cash flow of RMB 1.800 billion, down 44.08% YoY (Q8). With net profit of RMB 13.6 billion and operating cash flow of only RMB 1.8 billion, it looks bad on paper. Management's explanation: first-half operating cash outflows were "mostly for purchasing raw materials," partly purchases at the normal delivery cadence, plus "a large amount of advance stocking." The corresponding line items are prepayments and other long-term assets. This is a deliberate choice to trade cash for capacity, not a failure to collect. Accounts receivable rose from RMB 6.277 billion to RMB 15.001 billion, but management noted in Q2 that "payment terms and collection cadence have not changed for any customer"; the high balance is simply because June shipments were concentrated at month-end.

Financial expenses of RMB 562 million, up more than 15-fold YoY (Q13). The main cause was FX losses: about RMB 300 million in Q1 and about RMB 400 million in Q2. Management stressed that "most of this is an accounting loss at the reporting level, not an actual loss," including a portion arising from related-party transactions that is purely an artifact of accounting standards. Forward hedging is currently expensive (US dollar rates are high), so the company only hedges capex and FX conversion needs. Going forward, it plans to "minimize intercompany balances between subsidiaries to eliminate the accounting impact."

In plain English: this is a company with 95% of revenue overseas and large US dollar holdings, so exchange-rate moves get amplified on the income statement, but the actual cash loss is far smaller.

Investing cash outflow of RMB 6.816 billion (only RMB 696 million a year earlier), with construction in progress up from RMB 1.422 billion to RMB 3.707 billion (Q9, Q10, Q17). This is the most important number in the entire article. Management said several plants are being expanded, with "a small portion expected to be delivered in 2H26 and most expected in 2027," and that capex "will mostly begin converting into revenue in 2027." Capex will rise further in the second half, mainly for construction (equipment is generally completed in the first three quarters), and "capex will be relatively large from the second half of this year through next year."

Where does the firepower come from? The July 30 Hong Kong listing. The company priced its shares at HK$980 and listed on HKEX (03308.HK), raising HK$53.41 billion, the largest Hong Kong IPO of 2026, with 29 cornerstone investors including Hillhouse, BlackRock, a Temasek-affiliated entity, J.P. Morgan Asset Management, Alibaba and Tencent. Management was frank in Q17: "This year's Hong Kong offering was done because we foresaw that downstream customer demand will remain very strong, so we need to invest heavily in capex." The proceeds are allocated 35% to R&D, 30% to global capacity expansion, 15% to strategic acquisitions along the supply chain, 10% to supply chain resilience and 10% to working capital.

We discussed the industry significance of China's top three optical module makers all raising money in Hong Kong in TSMC Says Packaging Is Holding Customers Back, While China's Big Three Raised US$14 Billion: Two Faces of One Week. Now the interim report shows where that money is going.

7. Risks and Counterarguments

Will the competitive landscape loosen? (Q11) Management acknowledged: "Global industry demand will grow rapidly in 2027, so new competitors will enter and win some share." But it distinguished between two markets: CSP customers "still allocate the main 1.6T share to incumbent suppliers," so share is stable; new cloud customers (neoclouds, large-model companies) "may see more new entrants". It is a precise split: existing customers can be defended, while incremental customers are open to competition.

Peer comparison. All three Chinese optical module makers have reported 2026 first-half results: Innolight revenue RMB 41.778 billion / net profit RMB 13.651 billion (+182.5% / +241.7%), Eoptolink revenue RMB 20.910 billion / net profit RMB 7.529 billion (+100.3% / +91.0%), and TFC Communication revenue RMB 2.828 billion / net profit RMB 1.204 billion (+15.2% / +33.9%). Innolight has pulled ahead on both absolute scale and growth, but that also means the base is already very high. Sustaining the same growth rate in 2027 will require a very large absolute increase.

Valuation and sentiment. On August 24, all three stocks fell together, with combined turnover of RMB 71.4 billion. Strong results and a falling share price usually signal that "the good news is priced in" rather than weakening fundamentals, but it is also a reminder: this stock is priced on "2027 order visibility," so any signal of a 2027 capex cut will have an outsized impact.

The lack of guidance is itself a risk. The call gave no quantitative guidance on revenue or gross margin, and the company explicitly stated that the minutes do not represent a profit forecast. All expectations for 2027 currently rest on the qualitative statement that "customers have placed orders."

8. Conclusion

The easiest way to misread Innolight's interim report is to be dazzled by its numbers: beating the previous full year in six months means any metric could be taken as evidence of a peak. But read the call question by question and you find that management talked almost entirely about 2027: 2027 orders, 2.4T/NPO/XPO in volume production in 2H27, capex that will only turn into revenue in 2027, and product mix improvement that will only show up in 2027.

The real meaning of this half is that "this year's profits" and "the HK$53.4 billion raised this year" are being converted entirely into next year's capacity and materials. Operating cash flow down 40%, inventory up RMB 7.1 billion, construction in progress up 2.6x: none of this is loss of control. They are three sides of the same decision.

Three metrics to track next quarter:

1. Can gross margin hold at 46%–47%? Management's internal goal is "a slight increase"; a drop below 45% would mean upstream price increases are passing through more than expected.

2. Do Q3 inventory and operating cash flow improve QoQ? Management said "a QoQ improvement is expected in Q3," a direct test of whether the stocking cadence is normalizing.

3. The customer list and timing of the first NPO deliveries. Being "among the first suppliers to deliver NPO in 2027" is the hardest commitment of the call and the foundation of the 2027 gross margin story.

In one sentence: this is not a company harvesting a boom; it is a company putting every dollar the boom has earned back on the table, betting that optical interconnect will double again in 2027. The bet is placed; the cards won't be turned over until the second half of next year.