Innolight (300308) Report Interpretation

Overview

Goldman Sachs reinstates coverage of Innolight’s A shares and initiates H-share coverage at Buy. Its central case is that Innolight’s leading position in high-speed silicon-photonics optical modules will allow it to benefit disproportionately from AI-driven migration to 1.6T and 3.2T networking, while new scale-up and scale-across products add growth avenues.

Core views

Goldman Sachs characterizes Innolight as the largest global optical-interconnect provider by relevant 2025 revenue, with a 21.2% market share according to LightCounting and CIC. The report attributes the position to R&D, customer relationships, scale-related cost advantages, and production and supply-chain capabilities. It argues that the company’s early lead in silicon photonics (SiPh), including scaled commercialization of high-speed SiPh transceivers, makes it well positioned as AI infrastructure raises bandwidth and latency requirements. The report’s industry thesis is a rapid optical-specification upgrade. It forecasts the global optical-transceiver market at US$68bn in 2026E, US$131bn in 2027E and US$148bn in 2028E. The 800G-and-above segment is expected to grow at a 69% CAGR, reaching US$45bn, US$108bn and US$130bn, respectively. Goldman Sachs expects 800G shipments of 45m units in 2026E and 49m in both 2027E and 2028E; 1.6T shipments are projected to rise from 33m units to 71m and then 55m, while 3.2T shipments begin at 23m units in 2027E and reach 68m in 2028E. It expects silicon photonics to account for 60% of 800G, 80% of 1.6T and 80% of 3.2T transceivers over 2026-28E. The demand mechanism centers on AI-server networking. Goldman Sachs notes that NVIDIA’s GB200 uses 400G at the GPU layer and 800G through leaf, spine and core layers; GB300 shifts to 800G at the GPU layer and 1.6T at leaf and spine layers. Rubin and Rubin Ultra are expected to advance toward 1.6T and 3.2T. The report forecasts NVIDIA-powered rack-level AI servers at 50k, 92k and 148k in 2026-28E, and AMD-powered servers at 5k, 13k and 15k units. This migration, together with Google ASIC models using more 1.6T, underpins Goldman Sachs’ expectation of accelerated high-speed-module adoption. For Innolight, Goldman Sachs forecasts a 78% revenue CAGR for 2026-28E. It expects SiPh-module shipments to grow at a 71% CAGR and SiPh-module ASPs at an 8% CAGR as 1.6T and 3.2T become a larger mix. The company is also expected to expand from scale-out into scale-up and scale-across through optical engines, FAU, ELS modules and optical circuit switches. Despite capacity expansion, utilization is projected to remain elevated at 74-84% in 2026-28E. The report expects gross margin to improve from 42% in 2025 to 50% in 2028E as SiPh shipment contribution rises from 50% to 86% and as higher-bandwidth scale-up products carry better margins. Goldman Sachs forecasts revenue of Rmb131,869.5m in 2026E, Rmb276,712.1m in 2027E and Rmb417,779.1m in 2028E. It projects net income of Rmb40,432.1m, Rmb86,903.4m and Rmb137,110.3m, respectively, implying an 84% CAGR from 2026E to 2028E and net-margin expansion from 31% to 33%. Its 2026E and 2027E net-income forecasts are 25% and 42% above Bloomberg consensus, primarily because it assumes higher revenue and gross margin. The report sees 2H26 revenue rising 116% half-on-half, versus the 47% implied by consensus. The report also expects improving financial efficiency despite sustained investment. It forecasts ROE of 55-57% in 2026-28E, cash conversion cycle improvement from 122 days in 2025 to 80 days in 2028E, and a net-cash position through 2028E, with net debt-to-equity reaching -68%. Capex is forecast to rise 253% year-on-year in 2026E and remain high through 2028E, but Goldman Sachs expects free cash flow to stay positive and grow as operating cash flow expands. On competition and geopolitical risk, Goldman Sachs argues that rapid technology migration, tight raw-material supply and highly customized AI-server configurations make it difficult for cloud-service-provider customers to diversify away from established suppliers. It takes no view on the likelihood or outcome of reported potential US restrictions on imports of Chinese optical modules, but believes Innolight’s customer and supply-chain experience may help it adapt. The report also considers CPO competition manageable: even at 30% scale-out CPO penetration, it estimates pluggable-module value TAM would remain 10 times larger from GB300 to Rubin Ultra. Innolight’s offerings span pluggable modules, NPO engines and CPO engines. Goldman Sachs values Innolight A shares using 35.6x 2027E P/E, based on the relationship between earnings growth plus operating margin and peer valuation multiples. The multiple reflects 58% projected 2028E net-income growth, 43% 2028E operating margin and a low-end peer ratio of 0.35x, selected to account for concerns about CPO and new entrants. This produces the Rmb2,645 12-month A-share target. For H shares, it applies a 13% H-A premium based on the 60-day average for comparable technology peers, yielding HK$3,267. The H-share target implies 40.3x 2027E P/E versus the stated current 11.5x, reflecting the report’s expectation of a datacom-driven re-rating.

Analysis framework

Goldman Sachs combines a top-down forecast for AI-driven optical-transceiver demand and technology migration with Innolight’s market position, product mix, production capability and supply-chain analysis. It then models revenue, margins, cash flow and balance-sheet outcomes, compares forecasts with Bloomberg consensus, and applies peer-based forward P/E valuation with an H-A premium for the H-share target.

Methodology notes

• Industry AnalysisSupply-demand framework

  AI-server-driven demand and supply-chain analysis for high-speed optical transceivers.

  The report links AI-server rack growth and networking-specification migration to demand for 800G, 1.6T and 3.2T modules, while considering capacity, utilization and component availability.

• Industry AnalysisVolume-price decomposition

  Shipment and average-selling-price drivers of Innolight revenue.

  Goldman Sachs separates projected SiPh shipment growth from ASP gains resulting from higher-speed product mix.

• Valuation methodsP/E and PEG Valuation

  Forward P/E and PEG-based peer valuation.

  The A-share target applies 35.6x 2027E P/E, while the H-share target uses a 13% H-A premium and assesses the implied P/E and PEG against peers.

• Corporate Fundamentals and FinanceFree cash flow analysis

  Cash-flow analysis alongside capacity-expansion capex.

  The report assesses whether growing operating cash flow can keep free cash flow positive despite elevated capex through 2028E.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Innolight A shares (300308.SZ) and H shares (03308.HK)

  Primary covered securities expected to benefit from AI-driven high-speed optical-transceiver upgrades.

  Strengths

  Leading global position, 21.2% 2025 market share, early SiPh commercialization, high-speed product breadth, integrated production and diversified supply-chain capabilities.

  Comparison

  The A-share valuation uses peer multiples; the H-share target uses a 13% H-A premium based on comparable technology peers’ 60-day average.

  Risks

  Slower 800G+ demand or product ramps, market-share normalization, geopolitical restrictions and component shortages could reduce earnings estimates.

Key data

• A-share 12-month target priceRmb2,645.00Buy rating; versus stated price of Rmb814.00 and 224.9% upside.
• H-share 12-month target priceHK$3,267.00Buy rating; versus stated price of HK$1,006.00 and 224.8% upside.
• Revenue CAGR78%Goldman Sachs forecast for 2026-28E.
• Net-income CAGR84%Goldman Sachs forecast for 2026E-28E.
• 2026E / 2027E net-income forecast versus consensus+25% / +42%Above consensus, driven mainly by higher revenue and gross-margin assumptions.
• Gross margin42% in 2025 to 50% in 2028EExpected benefit from higher SiPh mix and scale-up expansion.
• Global optical-transceiver marketUS$68bn / US$131bn / US$148bnGoldman Sachs forecasts for 2026E / 2027E / 2028E.
• 2027E target P/E for A shares35.6xApplied to Innolight’s 2027E net income.

Impact & implications

The report argues that Innolight can translate the AI infrastructure cycle into faster shipment growth, richer product mix and margin expansion because of its SiPh lead and ability to serve scale-out, scale-up and scale-across networking. Its above-consensus earnings estimates and target multiples rest on the view that technology migration and supply constraints reinforce, rather than weaken, the position of established leaders.

Risks

• 800G, 1.6T and 3.2T optical-module demand may ramp more slowly than expected, reducing earnings estimates.
• Expansion of new optical devices and scale-out or scale-across opportunities may be slower than expected.
• Fiercer competition could cause Innolight’s 800G+ market share to normalize faster than expected.
• Potential US restrictions on Chinese optical-module imports and broader geopolitical tensions could pressure estimates and valuation.
• Shortages of upstream components such as EMLs, CW lasers and DSPs could constrain shipment growth.

What to watch

• The pace of 1.6T and 3.2T module ramp-up and adoption on new AI platforms including Rubin and Rubin Ultra.
• AI-server rack growth and the associated demand for high-speed optical networking.
• Innolight’s expansion into scale-up and scale-across products, including NPO, CPO, optical engines and switches.
• Supply availability for key upstream optical components and the company’s ability to maintain high utilization.
• Developments regarding potential US restrictions on Chinese-made optical modules.