Innolight Technology
Design, packaging, test and sale of high-speed optical transceiver modules (400G/800G/1.6T) plus emerging silicon-photonics/CPO; volume hardware supplier to hyperscaler and networking-OEM datacenter build-outs, largely make-to-order against customer capex cycles.
Earnings, margins, COGS & capex
Explosive AI-driven growth: FY2025 revenue CN¥38.24B (+60.25%) and net profit CN¥10.80B (+108.78%), with margins expanding as 800G and early 1.6T mix rose. Momentum accelerated sharply into 2026 - Q1 2026 revenue CN¥19.50B (+192.12% YoY) and net profit CN¥5.73B (+262.28% YoY), a single quarter already exceeding full-year 2024 net profit, implying a TTM run-rate well above the FY2025 base. Growth is concentrated in high-speed AI interconnect and increasingly in silicon-photonics/CPO.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~57¢ is cost of goods and ~5¢ operating expense, leaving ~38¢ of operating profit (~28¢ net).
Revenue trend
Margins
up ~8pp YoY on richer 800G/1.6T mix
expanding
up; net profit +108.78% on revenue +60.25%
COGS structure
Dominated by optical/electronic components - EML and DFB laser chips, DSP/retimer ICs, silicon-photonics engines, connectors and packaging materials; margin leverage comes from higher-speed mix (1.6T > 800G > 400G) and rising in-house silicon-photonics content that displaces bought-in laser arrays.
Capex
Capacity expansion in China plus a Thailand plant (TeraHop) with >$500M of announced capital increases for tariff/geographic de-risking; the frequently cited '~60% of capacity in Thailand' is not confirmed in primary disclosure and ramp progress has been questioned. Management targets scaling toward ~100,000 1.6T modules/month by end-2026 (a target, not disclosed capex); FY2025 module capacity ~28.06M units, production ~23.76M, sales ~21.09M.
Latest earnings
Strong upside surprise - revenue +192% and net profit +262% YoY, well ahead of prior run-rate
Management guiding to ~100,000 1.6T modules/month by end-2026 and rising silicon-photonics mix; 3.2T CPO mass production targeted 2027 (no formal numeric revenue guidance disclosed)
- Q1 2026 revenue
- CN¥19.50B (+192.12% YoY)
- Q1 2026 net profit
- CN¥5.73B (+262.28% YoY)
- FY2025 net profit
- CN¥10.80B (+108.78% YoY)
- Top-5 customer concentration
- ~76% of revenue (75.98%)
Growth drivers
- Hyperscaler + NVIDIA-platform AI datacenter capex driving 800G/1.6T transceiver demand
- Speed transition 400G -> 800G -> 1.6T lifting ASPs and gross margin
- Silicon-photonics ramp and NVIDIA-certified 1.6T SiPh modules (management cites >50% SiPh mix in high-speed volume)
- CPO roadmap: 1.6T CPO and a 3.2T CPO prototype targeting 2027 mass production (management roadmap, not verified yield/power)
- Thailand capacity intended to enable overseas (US hyperscaler) delivery despite trade barriers
Bull & bear
The best-positioned volume supplier of one of the most supply-constrained components in the AI datacenter - optical interconnect - with a widening 1.6T/CPO lead, expanding margins, and a Thailand footprint intended to keep it selling to US hyperscalers.
- AI interconnect demand is structurally short; Innolight has the yield and capacity lead at 800G/1.6T
- Q1 2026 (+192% rev / +262% profit) shows the 1.6T ramp is real and accelerating, not a peak
- Silicon-photonics/CPO integration turns a components-buyer into a margin-capturing platform - GM already +8pp
- NVIDIA certification + design-in creates switching-cost stickiness at the highest-value node
- Net-cash balance sheet and dividends fund capacity without dilution during hypergrowth
A single-market, customer-concentrated hardware vendor priced richly (~80x trailing / ~114x on FY2025 net income), structurally dependent on Western chips and hyperscaler capex, and only ownable via a China A-share.
- ~76% top-5 concentration means one hyperscaler capex pause craters the model
- Rich multiple (~$172B cap; ~82x TTM P/E, ~114x on CN¥10.8B FY2025 net income) - any deceleration de-rates hard
- Depends on Broadcom/Lumentum/Coherent/Marvell for lasers and DSPs - rivals that can prioritize their own modules
- Export-control / geopolitical risk could sever access to US customers or key components; Thailand ramp progress is questioned
- CPO/on-board optics and commoditization threaten the pluggable-transceiver moat over 3-5 years
- Not investable for US mandates; A-share liquidity, disclosure and FX add friction and risk
What it is worth
Multiple-based sanity check (P/E), triangulated against growth. Market cap ~$172B (CN¥1.23T) vs FY2025 net income CN¥10.8B implies ~114x on the FY2025 base; on a trailing-twelve-month basis (Q1 2026 replacing Q1 2025) net income is ~CN¥15B, so the reported trailing P/E is ~82x. Annualizing the Q1 2026 run-rate (CN¥5.73B/qtr -> ~CN¥23B) points to a forward P/E of ~54x if the 1.6T ramp holds.
A hyperscaler capex air-pocket or export-control shock stalls revenue and compresses margins; on a ~82x TTM (~114x FY2025) base the de-rate is severe (multiple + earnings both fall), with 40%+ downside plausible.
Growth decelerates from Q1's +192% toward a still-strong 30-50% as 1.6T matures; forward P/E compresses into the ~45-60x range and the stock roughly tracks earnings, with concentration/geo risk capping the multiple.
If 1.6T/CPO demand compounds and margins hold near ~28-30% net, forward earnings power (CN¥25B+) makes the ~82x TTM multiple look like ~45-50x forward - growth justifies the premium and the stock re-rates higher.
Valuation hinges almost entirely on whether AI-interconnect hypergrowth persists and margins hold. As a China A-share it is context/analysis only, never a US buy/own recommendation. Not financial advice.
SWOT
Strengths
- #1 global optical-transceiver maker by revenue/volume; scale and yield lead at 800G/1.6T
- Deep NVIDIA relationship - certified 1.6T silicon-photonics modules, design-in on AI reference platforms
- Rapid margin expansion (GM ~42.6%, +8pp) as high-speed mix rises
- Strong balance sheet, net cash, dividend-paying despite hypergrowth (CN¥1.11B FY2025 dividend)
- Thailand capacity intended to de-risk US-facing supply from China trade friction
Weaknesses
- Severe customer concentration - top-5 ~76% of revenue
- Component dependence on Western laser/DSP suppliers (Broadcom, Lumentum, Coherent, Marvell) who are also rivals
- Single end-market exposure: AI/datacenter interconnect capex cycle
- China-listed A-share - inaccessible to most US investors, geopolitical/export-control overhang
- Hardware commodity risk longer-term as speeds standardize and Chinese peers scale
Opportunities
- 1.6T ramp and 3.2T CPO (2027 target) extend the ASP/mix uplift
- Silicon-photonics vertical integration displaces bought-in laser arrays, lifting margin and supply control
- Co-packaged optics (CPO) transition where it claims certified, high-yield product
- Sovereign/enterprise AI buildouts widening the customer base beyond top hyperscalers
- Overseas (Thailand) footprint opening tariff-protected share of US demand
Threats
- Hyperscaler/NVIDIA AI capex digestion or air-pocket would hit revenue hard (historical precedent of ~50% revenue declines in prior corrections)
- US export controls / entity-list risk on advanced components or on shipping to US customers
- CPO and on-board optics could eventually disintermediate pluggable transceivers
- Intensifying competition from Eoptolink, Coherent, Lumentum and low-cost Chinese peers
- Laser/DSP chip supply constraints or allocation shifts by dual-role suppliers
Moats, dependencies & bottlenecks
Moats
World's largest volume plus claimed high 1.6T CPO yield is hard to replicate quickly, but speed nodes reset every ~2 years.
Certified 1.6T SiPh modules create switching costs per platform generation; must be re-won each cycle.
Moderate/rising Rising SiPh mix reduces dependence on bought-in lasers and lifts margin; deepening into CPO.
Intended to enable tariff-protected US delivery; capacity share not confirmed and replicable by well-funded peers.
Dependencies
Top-5 ~76% of revenue; revenue tracks a single capex cycle.
Supply / component Rivals double as suppliers; allocation or export limits would bite.
Supply / component Critical for pluggable modules; concentrated supplier base.
Regulatory / geopolitical Entity-list or tariff escalation could cut US-customer access or component inflow.
Must re-qualify each generation (1.6T -> 3.2T CPO).
Advantages
- Global #1 optical-transceiver revenue and volume position
- Early-mover, NVIDIA-certified 1.6T silicon-photonics module
- Rapidly expanding margins on high-speed mix (GM ~42.6%)
- Net-cash, dividend-paying balance sheet funding the ramp
- Thailand capacity as a tariff/geo hedge for US demand
Weaknesses
- Extreme customer concentration (~76% top-5)
- Single end-market (AI/datacenter interconnect) cyclicality
- Component dependence on rival Western chipmakers
- China A-share - inaccessible to most US investors, geopolitical overhang
- Commodity-hardware margin risk as speeds standardize
Bottlenecks
- Laser-chip (EML) and DSP allocation from a small set of Western suppliers
- 1.6T/CPO manufacturing capacity ramp and yield at scale
- Skilled photonics packaging/test labor and equipment for CPO
- Geopolitical clearance to keep shipping to US hyperscalers
- Working-capital intensity of building capacity ahead of demand
Top signals & trends
Top signals
1.6T ramp accelerating well beyond FY2025 base.
Mix shift to 800G/1.6T and SiPh integration is real margin, not just volume.
Design-in at the highest-value node.
Concentration risk if any hyperscaler pauses.
De-risks tariffs but signals how central US-facing demand and trade friction are.
Priced for continued hypergrowth; fragile to any miss.
Trends
Primary revenue driver; every AI cluster needs dense high-speed optics.
Each node lifts ASP and margin; Innolight leads the curve.
Margin/moat upside if it owns the transition; disintermediation risk to pluggables if others do.
Structural overhang on components and US-customer access.
Price competition risk as domestic rivals scale high-speed volume.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
DSP/retimer ICs and laser/PHY components.
Optical DSPs for high-speed modules.
EML/DFB laser chips (also a competitor).
Laser chips and photonics components (also a competitor).
Transceiver DSP/CDR and analog interconnect ICs.
Certification/design-in partner and demand driver; certified Innolight 1.6T SiPh modules (not confirmed as a named direct top-5 customer in disclosure).
Hyperscaler datacenter interconnect buyer (inferred top-5).
Hyperscaler cloud/AI fabric buyer (inferred top-5).
Azure AI datacenter interconnect buyer (inferred top-5).
AI cluster interconnect buyer (inferred top-5).
Switching OEM integrating high-speed optics.
Networking OEM customer.
US datacom-transceiver + laser-chip leader; both a rival and an upstream laser supplier.
US laser/photonics supplier expanding into datacom transceivers; supplier-and-rival dynamic.
Closest Chinese peer in 800G/1.6T; direct volume competitor (context only, not a buy call).
Pushes CPO/on-board optics and supplies DSPs/lasers - potential disintermediator of pluggables.
Chinese optical-component and transceiver maker; broad but less AI-concentrated (context only).
Optical DSP leader; enables rivals and pushes custom/CPO interconnect.
Contract optics manufacturer that assembles for many competitors; ecosystem rival on capacity.
Additional Chinese/global transceiver supplier competing on price at 400G/800G (context only).