The Zhitong Finance App learned that China Galaxy Securities released a research report saying that AI computing power is driving optical interconnection upgrades, and the entire phosphating industry chain is moving from “material bottlenecks” to “value revaluation”: with the large-scale deployment of 800G optical modules and the accelerated introduction of 1.6T into the next-generation AI cluster, phosphorylation (InP) is the only direct bandgap III-V material that can support 1310/1550nm long-range high-speed single-mode transmission. Festival”. In 2025, the global demand for InP devices is about 2 million, and the actual production capacity is only about 600,000 pieces. The gap between supply and demand exceeds 50%, and the production expansion cycle is long, yield is slow, and upstream resources are confined to zinc ore, so supply elasticity is extremely weak.
The main views of China Galaxy Securities are as follows:
The “association+regulation” of upstream steel resources is double restrained, and the purity bottleneck is stuck in the industrial chain
90% of the world's raw materials come from zinc ore by-products, and China's refining production accounts for about 70% of the world's refining production; after the Ministry of Commerce implemented export controls on materials such as phosphorylation in February 2025, prices were decoupled from east to west, and purification of high-purity steel above grade 6N (electrolytic vacuum distillation - multi-stage regional melting process) further limited the availability of raw materials at the substrate end, and upstream scarcity was transmitted to the midstream.
The midstream substrate oligopoly pattern is stable, and the 6-inch upgrade+production expansion cycle is the core
The substrate market share is smaller, but its technical barriers and market concentration are much higher than epitaxial: substrate manufacturing is a crystallization process from 0 to 1. The core difficulties are: 1) the core bottleneck is upstream, and substrate manufacturing relies heavily on ultra-high purity steel of grade 6N (99.9999%); 2) the long crystal process is extremely difficult. InP single crystal growth requires extreme environmental control at high temperature and pressure, and the expansion cycle is long, equipment delivery and yield are extremely slow; 3) High oligopoly: Global InP substrate Sumitomo (42%), AXT/Beijing Tongmei (36%), and JX Nippon Mining (13%) Total monopoly > 90% share. The Sumitomo Osaka Plant is expected to triple production, and AXT will raise US$632.5 million to expand production. It is expected that the gap will be filled. Epitaxial, on the other hand, grows a film on a substrate. Compared to substrates, it has a lower market concentration and a higher market share. Therefore, the bank believes that phosphating substrates, as the upstream manufacturing link in the industrial chain, currently have the highest technical barriers and the tightest supply and demand, and are also the most stranded link in the current AI optical module chain.
The shortage of downstream EML is forcing technology path differentiation, and InP is becoming irreplaceable on the light source side
Optical chips evolved along EML (current) → CW+SiPh (rapid development) → CPO+ELS (2026-2028) → silicon-based monolithic III-V (2030+); after EML was locked in production capacity and delivery by NVIDIA until 2027, the 800G supply and demand gap was 40-60%, driving CW+SiPh to accelerate penetration, but CWINP's base still depends on Sumitomo/AXT, and InP demand has not decreased but increased.
Industrial chain profits are concentrated on “substrate/epitaxial + high gloss chip”, and vertically integrated manufacturers expand their bargaining power
Substrate technology barriers and manufacturer concentration are the highest; epitaxial extension is scattered because MOCVD can be purchased, but the degree of customization is high and customer stickiness is strong; downstream EML/CW/UHP chips and CPOELS module ASP are 2-3 times more elastic. IDM (Lumentum, Coherent) and substrate material leaders (Sumitomo, AXT/Tongmei), which have full stack capabilities of InP long crystal+epitaxial + chip+module, will enjoy a gap premium.
Investment advice
The industry is in short supply and booming. It is recommended to focus on relevant leading targets, such as AXT (AXTI), Japan Sumitomo Electric (5802.T), IQE (IQE.L), AppliedOptoElectronics (AAOI.O), Lumentum (LITE), Coherent (COHR), and Ciena (CIEN.N), etc., and focus on the upward trend in profitability and valuation.
Risk Alerts
1. Risk of demand fluctuations due to CSP capital expenditure falling short of expectations; 2. Risk of disruptions in upstream supply and export controls; 3. Risk of substrate epitaxial expansion and yield falling short of expectations; 4. Risk of technology path substitution and CPO penetration falling short of expectations.