Earnings Call Highlights: Largan | Q2 2026 — First CPO Order Lands, but the Real Signal Is “GC Isn't Going Away”

2 days ago

9 min read

At Largan's (3008) July 9 earnings call, roughly 80% of analysts' questions were about CPO, not smartphone lenses. Three things set the tone. First, its Fiber Array (FA) won a first order from a volume-production customer: samples ship this month, a pilot line is due by the end of Q3, and volume ramp could come as early as mid-2027. Second, facing market panic over Corning's Glass Bridge, 林恩平 didn't play defense; he drew a line between technology routes: Glass Bridge targets EC, Largan stands on GC, and the two will coexist. Third, quarterly net income rose 352% YoY, but that figure is misleading: the year-ago quarter was weighed down by an NT$4.22 billion FX loss. What really matters is that operating income grew only 5.5% YoY and gross margin fell 4.2 percentage points.

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

1. Executive Takeaway

Signal one: CPO went from “evaluation” to “order” in just three months.

At the April earnings call, 林恩平 still said “FAU sampling and feasibility evaluation are under way, dB values have not met target, and volume production needs 1–2 years of line build-out.” On July 9 that became “we have one volume-production customer, samples ship this month, the customer will qualify as fast as possible, the pilot line will be done by end-Q3, and volume production could come as early as the middle of next year.” Largan only entered CPO as a latecomer last September and won its first order within ten months.

Signal two: Glass Bridge didn't hit Largan, because they are not on the same technology route at all.

林恩平's distinction is clean: Corning's Glass Bridge addresses Edge Coupling (EC), while Largan does Grating Coupling (GC). The GC market is currently clearly larger than EC, “and EC can't cross over into GC either,” so the two will coexist long term. This is a technical judgment, not reassurance, but it also reveals the direction Largan is betting on.

Signal three: the profit surge is an accounting phenomenon, not an operating one.

Q2 net income of NT$4.67 billion, up 352% YoY, looks like a turnaround story. But operating income for the same quarter was NT$5.148 billion, up just 5.5% YoY, and gross margin was 49.4%, down 4.2 percentage points YoY. The entire gap is non-operating: the historic NT$4.22 billion FX loss in Q2 2025. The core smartphone lens business is actually moving slowly.

2. Revenue and Financials

Q2 2026 consolidated revenue was NT$13.665 billion, down 12.1% QoQ and up 17.1% YoY. Gross margin was 49.4%, flat QoQ and down 4.2 percentage points YoY. Operating income was NT$5.148 billion, an operating margin of 37.7%, up 5.5% YoY. Net income was NT$4.67 billion, up 352% YoY. Quarterly EPS was NT$35.72. Quarterly capex was about NT$2.78 billion and depreciation about NT$2.1 billion.

First-half cumulative: consolidated revenue NT$29.209 billion, gross margin 49.4%, net income NT$10.793 billion (up 44% YoY), EPS NT$82.35. The company did not provide next-quarter guidance figures.

Here is something easy to misread: the 352% net income growth rate has no operating meaning. In Q2 2025, with the New Taiwan dollar appreciating nearly 11% in a single quarter, Largan booked an NT$4.22 billion FX loss and quarterly EPS of only NT$7.73, its worst quarter in 13 years. Measured against a collapsed base, the growth rate is distorted. Switch the base to operating income and growth immediately drops to 5.5%.

Gross margin deserves more attention. Largan once held margins “in the 50s” for four straight quarters; full-year 2025 gross margin was 50.37%, Q4 2025 fell to 47.8%, and both of the first two quarters of 2026 were 49.4%. The “50s” era has been over for three quarters.

Largan Q2 2026 core financials (revenue / gross margin / operating margin / net income / EPS / YoY / QoQ / first-half cumulative). Image source: Simple Tech Trend | Data source: Largan financials announced at the 2026/07/09 earnings call (unaudited)

Q2 product mix remained mid-range: 10MP at 50%–60%, 5MP at 20%–30%, 20MP and above at 10%–20%, and 8MP below 10%. The high-end share didn't jump noticeably, which also explains why gross margin can't climb.

3. Technology and Business Highlights: the FA moat isn't optics, it's tolerance management

Largan has picked a precise position in the CPO supply chain: it doesn't make finished FAUs, only high-precision FAs, which it sells to module-maker customers who assemble them into FAUs. 林恩平's logic: everyone wants to make FAUs, but the pain point is the lack of high-precision FAs; PMLAs and V-grooves aren't in short supply, FAs are. So selling FAs to everyone is a business model that competes with no one.

The real technical point is this line: “We use imperfect V-grooves with imperfect fibers to make perfect FAs.”

The traditional approach picks the best V-grooves and the best fibers, yet the stacked tolerance is still over 1 micron, forcing heavy rework and scrap, so yield can't bring costs down. Largan does the opposite: its in-house process combines two parts that each have tolerances into an assembly with better than 0.3-micron precision, versus about 0.5–0.8 microns for the strongest competitors. The technique is patent-pending.

The barrier for FAs isn't “lining fibers up neatly” but “lining them up neatly without binning parts.” The former is optics; the latter is manufacturing.

The second technical point is automation. Mature CPO automation equipment simply can't be bought on the market, so Largan builds most of its line equipment itself. This is the real asset it brings over from smartphone lenses: vertically integrated capability in molds, machines and precision assembly. The optical communications industry has long relied on labor-intensive assembly, which has no room to survive in Taiwan. That line is both an industry observation and Largan's statement of differentiation.

Worth watching is the number of stacked rows. The volume-production win is a 1-row design; 2-row is still at POC. At Computex Largan showed a four-row stacked, lidless structure; Japanese players (such as Sumitomo) use through-glass holes or add a lid to do two rows. The more rows, the clearer Largan's advantage, but that demand won't appear until AI compute moves up another one or two generations.

Why this step is the real bottleneck for CPO volume production, and the fifteen supply chain players competing for it, is covered in full in Can the Fiber Be Swapped? The Most Underrated Step in CPO Volume Production, and the 15 Companies Racing for It.

4. Management Outlook: from “uneasy” to “solid”

The most analyzable part of the call is the shifting gradient of certainty in 林恩平's wording.

April earnings call: FAU still in sampling and feasibility evaluation, dB values not at target, volume production needs 1–2 years of line build-out.

June shareholders' meeting: “At the last earnings call I still felt uneasy, but now that I see products coming out and the first automated line about to be set up, I feel more solid.”

July earnings call: one volume-production customer secured, samples ship this month, and the customer will complete qualification as fast as possible; pilot line done by end-Q3; the FA volume line could be built as early as mid-2027, with formal volume production once yield hits target.

In three months: “evaluation” → “confidence” → “order.” This kind of time compression is unusual in the optical communications supply chain, where qualifying a new supplier usually takes more than a year.

But note what 林恩平 didn't say:

Pricing isn't fully settled. He said plainly that FA ASP will be higher than smartphone lenses (high material cost), but “poor yield will also do more damage to profit,” and CPO gross margin can't be estimated yet.
Core business outlook is murky. July shipment momentum is better than June, and August better than July, but overall Q3 revenue is “hard to see,” and quarterly gross margin “depends on yield.” The number of projects from the main customer is about as previously flagged, just with schedules pushed out; if new phones sell poorly after launch, there could still be downward revisions.
Timing of CPO revenue contribution. The company only says volume revenue could start as early as mid-2027; analyst estimates put CPO at about 1%–2% of revenue in 2028. The gap between those two numbers is the space for market imagination.

Using STT's wording tiers: “one volume-production customer secured” is confirmed; “volume production as early as the middle of next year” is a target; “volume production starts once qualification goes smoothly” is conditional. The three must not be conflated.

5. Supply Chain and Customer Clues

Clue one: the GC camp vs. the EC camp sets Largan's ceiling.

Largan's share price came under pressure after Corning announced Glass Bridge, essentially because the market feared the EC route would swallow GC. 林恩平's technical distinction: EC couples light into the PIC from the chip's side facet; GC uses micro-structured gratings to guide light vertically in and out. Market observers note that NVIDIA GPUs mainly use GC, while AMD and Broadcom lean toward EC. If that mapping holds, the ceiling for Largan's FA is, to some degree, tied to the NVIDIA route.

GC's advantages are greater chip design flexibility, easier testing and alignment, and suitability for large-scale automated production; EC has extremely tight alignment tolerance, and a slight offset causes severe optical loss, which is exactly the problem Glass Bridge aims to solve. That means Glass Bridge isn't here to take Largan's orders, but to make the hard-to-mass-produce EC route manufacturable. Long term, that actually grows the market. How Corning's materials dominance is extending from packaging substrates to the fiber coupling layer is broken down in Corning GlassBridge: Glass Materials Dominance Extends from Packaging Substrates to the Fiber Coupling Layer.

Clue two: the customers are module makers, not chipmakers.

Largan said explicitly that FAs are mainly delivered to module-maker customers, who assemble them into FAUs themselves. That puts it further upstream in the supply chain, away from the GPU makers' qualification process, but it also means its volume depends on module makers' market share to scale.

Clue three: components are outsourced, which benefits upstream suppliers.

Components such as V-grooves are currently mostly purchased externally, with no in-house plans for now. That is a real order lead for Taiwanese precision optical component suppliers (such as Ability Opto-Electronics).

Clue four: capacity and expansion.

Largan currently has ten plants in Taichung plus one leased plant. 林恩平 admitted that both talent and land are short; if CPO succeeds, buying more land to expand isn't ruled out, and if demand grows beyond expectations, outsourcing part of production isn't ruled out either. The word “outsourcing” coming from Largan is itself a signal: the company has long been known for vertical integration and closed production lines, and for CPO it opened its lines to customer visits for the first time.

Clue five: structural pressure on the smartphone business.

林恩平 once said something weighty: the more AI advances, the less lens specs need upgrading, which amounts to “eating our future.” Variable aperture starts shipping in Q3, periscope lenses will upgrade to more elements in 2027, main cameras have new specs, and foldable-phone lenses are “not relatively more difficult.” These are all incremental, but none can support a second growth curve. CPO can.

6. Conclusion

The real content of this call isn't the financials. The financials tell you only one thing: gross margin in the smartphone lens business is in structural decline, operating income grew 5.5% YoY, and a threefold net income jump produced by an FX reversal is not strength.

The real content is that a company that has made plastic aspheric lenses for twenty years used “tolerance management” rather than “optical design” to break into the most constrained link in the CPO supply chain within ten months, and won an order. Its moat was never lenses; it is the manufacturing ability to assemble imperfect parts into perfect products, and that ability is transferable.

Three metrics to track next quarter:

Whether FA sample qualification passes in Q3. This is the only hard gate. Samples ship in July and the pilot line is done by end-Q3; if the Q3 earnings call (October) doesn't include the words “qualification passed,” the mid-2027 volume timeline will slip.
Pilot-line yield numbers. 林恩平 has already set expectations: material costs are high, and poor yield hurts profit more. Whether CPO gross margin can beat the 49.4% of smartphone lenses rests entirely on this.
Whether the 2-row design converts from POC to orders. Single-row FA is a red ocean; multi-row stacking is where Largan claims its advantage. When second-row orders appear will determine the slope of this curve.

As for Glass Bridge, it isn't Largan's rival; it is an accelerator for the EC route. The real worry isn't Corning, but whether GC's share gets squeezed in next-generation packaging architectures. That depends on NVIDIA, not Corning.