Earnings Call Highlights: Largan | Q3 2026 — Gross Margin Hits a 13-Year Low as Capex Goes All In on FAU

Oct 10

9 min read

• Core business is slipping; the books are holding up: revenue rose 14.6% QoQ, but gross margin fell to 42.2%, the lowest since Q1 2013, and operating income dropped about 8% QoQ. Net income still grew 11% QoQ, thanks to NT$1.357B in non-operating income and an effective tax rate that fell from about 23% to about 15%.

• Green power has lowered the gross margin ceiling by a notch: customer contracts require Largan to buy green power, 2027 pricing is already settled, and the cost can't be passed through until 2028 at the earliest. In 2027 the only lever left is yield.

• Largan's CPO role just got bigger: last quarter the line was "FA only, FAU undecided." This quarter the customer wants FAU plus carrier outright, and Largan has handed a newly completed fab, its second-largest site, entirely to FA/FAU, targeting capacity by end of Q2 2027. With handset lens visibility down to one month, capital allocation has already shifted to optical communications.

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

1. Three Core Signals

First, the gross margin problem has gone from "yield" to "yield plus cost structure." When Largan's margin slipped in the past, the reasons were usually FX, purchased components or new-model yield, all of which pass. This quarter added one that won't: green power. Chairman 林恩平 used the word "structural" himself.

Second, handset lens visibility is now one month. October shipments are similar to September, November starts to decline, and customers won't give long-term forecasts. The pull-in peak for the iPhone 18 variable-aperture lens has passed, and the customer started cutting in September.

Third, in CPO Largan is moving from "component supplier" to "sub-assembly contractor." Its FA (fiber array) has passed qualification, and the customer wants finished FAUs (fiber array units) shipped directly, mounted on a carrier. It is the first time since entering optical communications that Largan takes responsibility for an entire coupling assembly step.

2. Financials: Revenue Came Back, Gross Margin Didn't

Q3 revenue was NT$15.651B, up 14.6% QoQ and down 11.5% YoY. The YoY decline mainly reflects a different customer pull-in schedule this year, with peak-season volume less concentrated than last year.

• Gross margin 42.2%: down 7.2 points QoQ and 5.0 points YoY, the lowest since Q1 2013. The previous two quarters were both at 49.4%.

• Operating margin 30.3%: versus 37.7% in Q2, down 7.4 points; operating income was about NT$4.73B, down about 8% QoQ.

• Non-operating income NT$1.357B: including about NT$197M in FX gains; non-operating income made up 22% of pretax income.

• Net income NT$5.182B: up 11.0% QoQ, down 26.8% YoY.

• EPS NT$39.62: versus NT$35.72 in Q2 and NT$53.05 a year earlier.

Break the income statement apart and the structure is clear: pretax income of NT$6.092B was essentially flat with Q2's NT$6.099B. Net income rose only because the effective tax rate fell from about 23% to about 15%. In other words, the core business declined QoQ, and the reported growth came from non-operating income and tax.

Nine-month revenue was NT$44.86B, up 2.1% YoY; gross margin was 46.9%, 4.5 points below last year's 51.4%; operating income fell about 9% YoY. But nine-month non-operating income totaled NT$3.78B versus only NT$440M a year earlier, so net income of NT$15.976B still rose 9.8%, and EPS of NT$121.98 rose 11.8%. A large part of 2026 EPS growth is propped up by non-operating income, which should be stripped out when assessing the core business.

Largan Q3 2026 earnings data card: revenue, gross margin, operating margin, net income and EPS

3. Gross Margin Breakdown: Green Power Is Structural, Yield Is the Only Lever

林恩平 gave two reasons for the Q3 gross margin: weaker yield on the new model, and higher green power costs. The two are completely different in nature.

Yield is cyclical. A variable-aperture lens must focus correctly at multiple aperture settings, so it needs even more inspection stations than a periscope lens, and high early-ramp rejection is expected. Management called yield "still OK" and improving; the picture should be clearer by end of Q4 or Q1 next year.

Green power is structural. Customer contracts require buying a set share of green power, with the cost allocated by that customer's shipment share. Asked how many points it costs, the company only said the impact is "at least" meaningful and did not disclose a number. The timing matters more: 2027 product pricing is already settled. If prices included green power, margins could recover, but that won't happen before 2028.

In 2027, the only lever Largan has to improve gross margin is yield.

For Largan, this means the 50% to 60% gross margin range of the past decade-plus is being pushed down a notch by a cost unrelated to manufacturing capability. Even with excellent yield, margin can only move from "early ramp" back to a "post-green-power normal," not back to last year's level.

4. Handset Lenses: Visibility Down to One Month

Looking at Q4, 林恩平's signal was very short: October similar to September, decline starting in November, with no long-term forecasts from customers. Q4 utilization is still "all full," but that looks more like product mix and yield losses eating capacity than strong demand.

The iPhone 18 is this year's only major spec upgrade, adding variable aperture to the main camera for the first time, with Largan as the main supplier. But management's take: yield is fine, the customer's peak has passed, cuts started in September, and "this generation doesn't look like it's selling very well."

The spec upgrade cadence is also being pushed out:

• 2027: only a minor main-camera upgrade (sampling now).

• 2028: the next periscope upgrade, toward more lens elements and higher resolution. Last quarter the periscope upgrade was said to start "next year"; this quarter it moved to 2028.

Q3 shipment mix by resolution: 10MP-class at 60% to 70%, 20MP and above at 10% to 20%. The high-end share did not rise noticeably.

The handset lens story is simple: 2027 has no major upgrade to lift ASP, prices are locked, green power eats into margin, and visibility is one month. That explains why about 90% of the call's questions were about CPO.

5. CPO: From FA to FAU Plus Carrier

The biggest change this quarter is that Largan's scope in optical communications extended a full step downstream. Compare it with our Largan Q2 2026 earnings highlights: back then the line was "mainly delivering FA to module-maker customers," with components mostly purchased, and at the June shareholder meeting 林恩平 even said "whether to make FAU is still undecided."

Where things stand this quarter:

• FA qualification "should count as passed": one volume customer, pricing agreed; the FA pilot line was ready in September with equipment installed.

• The customer wants FAU directly. The flow: make the FA in-house → buy the PMLA (prism micro-lens array) → couple them into an FAU → mount the FAU on a carrier. The carrier is the board under the FAU with three alignment balls, and the board itself is purchased.

• The mid-2027 shipment form is "FAU plus carrier". The customer has seen the FAU demo line, and qualification of the new line will be discussed this month.

• Timeline: confirm capacity scale this month, build the first tranche by end of Q2 2027, with FAU volume production in 2H27.

Why jump from FA to FAU? 林恩平 was blunt: "It's not up to us." With separate suppliers, poor yield leads to finger-pointing; shipping a complete unit minimizes disputes. This is exactly what CPO Deep Dive 4/6: Fiber-to-Chip covered: loss and tolerance at the coupling step are the hardest place in the CPO chain to assign responsibility, so customers would rather have one supplier own everything from FA to carrier.

Capital allocation on capacity has clearly shifted too:

• A newly completed fab, roughly the size of its second-largest existing site, is fully dedicated to FA/FAU.

• Newly purchased land and buildings are also all planned for FA/FAU, with construction starting next year. Two completion dates are in circulation: our call notes say 2030, while media reports say usable as early as about 2028, possibly reflecting phased build-out. We'll check again next quarter.

• Q3 capex of NT$8.18B was mostly land; depreciation was NT$2.04B. The company expects 2027 capex to exceed 2026, saying it will "cost quite a lot," but it has no need to raise funds. Cash stood at about NT$110.9B at the end of Q3.

• In-house capacity may not be enough, so outsourcing part of the capacity is not ruled out, though every tool needs fine-tuning.

It's important to separate "becoming Largan's business" from "becoming Largan's profit." FA/FAU is a product with "a very high unit price, but a very high BOM too": fiber, PMLA and carrier are all purchased, and whether purchased materials count toward revenue isn't decided. Revenue could scale quickly, but gross margin depends entirely on yield, which won't be known until volume production.

6. What Largan Chooses Not to Do

More informative than what Largan will do were the boundaries management drew on its own:

• No ferrules or connectors: the customer doesn't want Largan making them, and Largan only targets FA and FAU, where "barriers to entry are especially high." That leaves MT ferrules and connectors to incumbents, and explains why Largan's path doesn't overlap with FOCI's FAU patents.

• No PMLA: the customer qualified another vendor long ago, so Largan must buy it. Customers currently prefer semiconductor-process PMLA; molded glass only has an edge in some configurations. The value sits in FA/FAU assembly, not in the molded glass Largan is best at.

• No worries about V-grooves: supply is ample, and Largan's active alignment (AA) process can turn even out-of-spec V-grooves into in-spec FAs.

• CPO vs NPO doesn't matter: the two differ only in packaging location, with little impact on FA or FAU themselves; specs are looser if NPO goes the VCSEL route. This matches our view in NPO Is Not CPO's Appetizer: the route battle mainly affects the optical engine and laser side, while coupling components win either way as long as they hold the high bar.

• Skeptical of microLED optical interconnect: 林恩平 sees it as too power-hungry and too large, "unlikely"; the value lies in hard-to-make single-mode FA, and he is cautious on multimode fiber.

• GC first, EC later: next year's volume product is GC (grating coupling), while EC (edge coupling) is still in co-development with the customer. FAU is moving toward a detachable design, which pushes more tolerance into the FAU itself and concentrates the difficulty in Largan's step.

7. Grading Management's Language by Certainty

Put the wording on different items side by side and the certainty levels become clear:

• Confirmed: FA pricing agreed, FA pilot line ready, 2027 handset lens pricing settled, Q4 utilization full, no fundraising needed.

• Near-confirmed: FA qualification "should count as passed," first capacity built by end of Q2 2027, shipment form "should be" FAU plus carrier.

• Still targets: FAU new-line qualification (talks only this month), FAU volume production in 2H, yield and gross margin ("depends on yield").

• Furthest out: multi-row FA in 2028 at the earliest, "not visible next year"; whether the line holds through 2030 is "unpredictable."

2027 can prove "we can make it"; only 2028 can prove "we can make it at scale."

8. Supply Chain Clues

• In 2028 the next generation goes straight to CSPs: in this generation Largan is a contractor with a single volume customer; for the next generation it is already talking directly with more than one CSP, mostly about multi-row FA. Moving from building for module makers to being named by end cloud operators is the shift most worth tracking mid-term.

• Beneficiaries of FAU purchased materials: PMLA (from the vendor the customer already qualified), carriers and alignment balls, and fiber are all purchased, so Largan's capacity ramp lifts demand in those segments.

• Demand for high-density fiber assembly really is growing: ECOC 2026: Corning fits 1,152 fibers into a 1U switch already showed that CPO's challenge is moving from the chip to the assembly line, consistent with the claim that FA and FAU are in short supply.

• In-house equipment: FA and FAU alignment tools are essentially all built in-house, with automation advantages in alignment precision, speed and inspection. It's one of the few capabilities Largan can carry straight over from handset lenses, and the reason the customer calls its line "the best we've seen."

9. Risks

• Double pressure on gross margin: handset-side green power can't be passed through until 2028, and optical-side high-BOM products depend on yield, so getting overall 2027 gross margin back above 50% won't be easy.

• Single-customer concentration: there is only one FA volume customer and the FAU new line isn't qualified yet; any qualification delay directly pushes back mid-2027 revenue contribution.

• Revenue recognition undecided: if purchased materials count as revenue, revenue will be inflated and gross margin diluted, and the market may misread the growth rate.

• Non-operating income isn't sustainable: NT$3.78B of nine-month non-operating income propped up net income growth; without a stronger core business, the 2027 EPS base is high.

• Shifting timelines: the periscope upgrade moved from 2027 to 2028 and the new fab's completion date has two versions, so management's long-range timing estimates deserve a discount.

10. Conclusion

Largan gave a very honest picture this quarter: the core handset lens margin is being re-anchored by a non-manufacturing factor like green power, while the company has bet its best fab and newly bought land on FAU. This isn't a second leg to diversify risk; it's a clear capital allocation pivot.

Three things to track next quarter:

• Whether Q4 gross margin stops falling and returns above 45%, to judge how much green power cost yield gains can offset.

• FAU new-line qualification results and capacity scale; the numbers confirmed with the customer this month will set the scale of 2H27 revenue.

• The accounting decision on whether purchased materials count as revenue, which directly affects how 2027 revenue and gross margin should be read.

In one sentence: Largan's window for pricing power in handset lenses has closed; in 2027 it has to prove that alignment precision can win pricing power back in optical communications.