Earnings Highlights: LandMark Optoelectronics (3081) | 2026 Q2 — 57% Gross Margin, Record EPS of NT$4.62, and Multi-Wavelength SiPh Lasers as a Moat

2 days ago

7 min read

This quarter LandMark didn't just grow revenue — its entire income statement moved up a level: quarterly revenue of NT$1.221 billion, up 121% YoY, with gross margin of 57%, net margin of 38%, and quarterly EPS of NT$4.62, all record highs. The real signal isn't the size of the numbers but the structure: Datacom (almost entirely CW lasers for silicon photonics) now accounts for more than 80% of revenue, and the company has committed nearly NT$6 billion of CAPEX across three quarters to a single goal — growing CW laser capacity 2.5x by 2027. While peers are still fighting a price war on single-wavelength parts, LandMark is using multi-wavelength capability to raise both the technical barrier and the ASP. That is the hardest bone underneath this report.

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

1. Three Signals to Remember First

First, profit quality improved across the board. Doubling revenue isn't unusual; what is unusual is gross margin climbing from 43% a year ago to 57% at the same time, with operating margin reaching 47%. This growth isn't volume bought with price cuts — the product mix is genuinely shifting toward higher-margin products.

Second, the growth momentum isn't stopping. Management guided 3Q26 to a sequential increase "similar to or even better than 2Q25" — 2Q25 grew +21.7% QoQ, so the company itself has drawn the floor for next quarter's growth around +20%, using the rather optimistic phrase "even better."

Third, this is a CAPEX bet, and the chips are already on the table. Cumulative CAPEX over the past three quarters is nearly NT$6 billion, targeting 2.5x current capacity in 2027 (YoY +150%), with the same pace planned for 2028. For an epitaxy house that is light on design and heavy on assets, this ties the next two years to one assumption: that silicon photonics laser demand will not break.

2. Revenue and Financials: The Whole Income Statement Moves Up a Level

Quarterly revenue was NT$1.221 billion, +35% QoQ and +121% YoY; gross profit was NT$700 million, with gross margin at 57% (+2 pts QoQ, +14 pts YoY). Operating expenses were NT$129 million (+6% QoQ, +42% YoY, mainly employee profit-sharing accrued in proportion to higher profits), and operating income was NT$573 million, with operating margin at 47% (+6 pts QoQ, +20 pts YoY). After roughly NT$10 million of non-operating income, net income was NT$468 million, a 38% net margin (+48% QoQ, +384% YoY), and quarterly EPS of NT$4.62 set an all-time high; first-half EPS totals roughly NT$7.75.

One item worth pulling out separately is depreciation: about NT$75 million this quarter, holding at a single-digit share of revenue. This matters — even as CAPEX rises sharply, as long as revenue scales alongside it, depreciation's dilution of gross margin stays limited. It is one reason gross margin could still rise this quarter.

3. Business Highlights: SiPh Takes 80%, and Multi-Wavelength Is the Real Moat

Start with the revenue mix. In 2Q26, Datacom was 80–85% of sales, the vast majority of it CW (continuous-wave) lasers for silicon photonics (SiPh); Telecom fell to 5–10%, and other products were about 5%. In other words, LandMark has gone from a company that "does a bit of every laser" to one that is "all-in on SiPh light sources."

The technology roadmap matches this structure. 1.6T products are already in volume production, while faster 3.2T and higher-power light sources are in co-development and validation with customers. The laser at the very top of a SiPh module — and the hardest part to make — is widely seen as the most fragile link in the whole CPO/SiPh supply chain. We broke down that bottleneck in The Weakest Link in CPO Is the Laser, and that is exactly the position LandMark occupies.

LandMark's real differentiation is multi-wavelength. Management said bluntly that nearly all peers claiming CW laser capability can only do single wavelength, while multi-wavelength parts are significantly harder to manufacture and command a higher price. That is why Chinese competitors, despite aggressive capacity expansion, will struggle to compete head-on with LandMark in the near term. We untangled the route debate between multi-wavelength sources, comb lasers and DFB arrays in Will Comb Lasers Replace DFB Arrays? Understanding CPO Light Sources.

As for why gross margin keeps rising, management's answer was candid: CW lasers have been in volume production for years and yields are mature, so further gains no longer come mainly from yield, but from a product mix shifting toward higher-margin parts, while the impact of rising substrate prices "has been fully reflected in the gross margin rising quarter by quarter." In plain terms: LandMark now has the pricing power to pass cost pressure on rather than absorb it.

4. Management Outlook: Two Years Bet on One Capacity Curve

Next quarter (3Q26): revenue growth QoQ similar to or better than 2Q25 (2Q25 QoQ was +21.7%), with SiPh shipments continuing to rise over the next several quarters.

Capacity: nearly NT$6 billion of cumulative CAPEX over the past three quarters will drive 2027 capacity to 2.5x current levels (YoY +150%), with the company hoping to sustain this rapid expansion pace in 2028. Expansion is concentrated in epitaxy; back-end processing after epitaxy remains outsourced.

Product roadmap: CW lasers remain the near-term focus. Only after 2028 will part of the capacity be extended to EML (electro-absorption modulated lasers) and even PD (photodetectors) — though neither is new (EML has been in volume production for about 20 years, and PD since day one). Management stressed that capacity is too tight in 2H26 to add any new product lines.

There's an interesting contrast here: LITE (Lumentum) argues that EML has the edge in 3.2T architectures, while LandMark responds that "EML and CW are not substitutes" — each will win in different scenarios across scale-out, scale-up and scale-in. Read this EML vs CW debate alongside Earnings Highlights: Lumentum (LITE) FY2026 Q4 — for the same 3.2T generation, the two leaders give different answers.

The depreciation target was also updated: management still aims to keep depreciation below 8–12% of revenue, stressing that this already accounts for all current and future CAPEX plans. Major customers are discussing demand on a 3–5 year horizon, and the company plans CAPEX on a 3+ year basis. This frames the risk of the "capacity bet" against long-dated customer commitments.

5. Supply Chain and Customer Clues: End Customers Are US CSPs, but Most Invoices Go to China

LandMark has two customer models. In the first, it co-develops directly with a CSP (cloud service provider); the CSP designates the downstream module maker, and LandMark delivers lasers to that module maker. In the second, the CSP lacks laser co-development capability and orders modules directly from module makers, which then buy chips or epi wafers from LandMark. Major CSPs in both the US and China are already shipping through module makers that work with LandMark.

Here is a signal that's easy to misread: revenue by region in the financial statements does not equal where the end CSP is located. Management noted that about 75% of global data center module capacity is in China, so even if 70–80% of LandMark's shipments go to China, the end customers may still be US CSPs — and by end customer, it is "definitely mainly US CSPs." This is an important lesson for interpreting the "China exposure" of Taiwanese optical companies. We clarified the debate over whether NVIDIA's shift from CPO to NPO hurts laser demand in Is NVIDIA's Shift from CPO to NPO Bad News for the Laser Supply Chain?; LandMark's view is that NPO and CPO make little difference to CW laser demand, and CPO is still unlikely to see large-scale commercial deployment in the near term.

Upstream, watch two things. InP (indium phosphide) substrates: LandMark works closely with major global substrate suppliers and sees no shortage risk over the next several quarters; most shipments now use 3-inch or larger wafers. As laser power rises and die sizes grow, 4-inch and larger substrates are more cost-effective, but 4-inch+ InP is a very small share of global supply — a hidden bottleneck across the supply chain. This observation that "the tightest link has moved all the way upstream to the laser" is the same story as Broadcom Extends Visibility to 2027, but the Hardest Bottleneck Is a Laser. Foundry side: back-end processing is outsourced, and the number of foundry partners keeps growing (existing partners' capacity falls far short of LandMark's needs). Whether LandMark ships epi wafers or chips depends on foundry capability and customer demand — not a ratio LandMark sets unilaterally.

6. Conclusion: A Report That States the Bet Clearly

LandMark's 2Q26 message was blunt: demand is strong, yields are mature, and the product mix is moving up-margin, so the company has put three quarters and NT$6 billion of CAPEX into CW laser epitaxy capacity, betting on 2.5x in 2027 and another wave in 2028. Multi-wavelength is its hardest moat today, while large-diameter InP substrate supply and foundry capacity are the two constraints it openly acknowledges.

Over the next three quarters, there are three metrics to watch: first, whether gross margin holds above 55% (the tug-of-war between product mix and substrate costs); second, whether 3Q26 revenue really grows ≥ +21.7% QoQ (testing how much substance lies behind management's optimistic wording); and third, equipment delivery progress toward 2.5x capacity in 2027 (lengthening lead times for MOCVD, e-beam and lithography tools are the biggest execution risk in this bet). The results prove demand is real; only one question remains — can capacity come online fast enough?