Artificial intelligence computing power is driving an upgrade in optical interconnect technology, pushing the entire indium phosphide (InP) industry chain from a "material bottleneck" towards a "value reassessment". With the large-scale deployment of 800G optical modules and the accelerated introduction of 1.6T, alongside the transition of Co-Packaged Optics/Silicon Photonics from samples to mass production in next-generation AI clusters, indium phosphide's strategic position has elevated from an "optical communication substrate" to a "critical bottleneck in AI infrastructure". This is due to its unique property among III-V materials as a direct bandgap semiconductor capable of supporting long-distance, high-speed single-mode emission at 1310/1550nm wavelengths.
China Galaxy Securities Co., Ltd. estimates global demand for InP devices will reach approximately 2 million units in 2025, while actual production capacity stands at only about 600,000 units. This results in a supply-demand gap exceeding 50%. Expanding capacity faces significant challenges: long lead times for expansion, slow yield ramp-ups, and a constrained upstream supply of indium, which is a byproduct of zinc mining, resulting in extremely low supply elasticity.
Upstream Constraints: Dual Pressures from Byproduct Dependency and Export Controls
Approximately 90% of the world's primary indium originates as a byproduct of zinc mining, with China accounting for roughly 70% of global refined production. Following the implementation of export controls on materials including indium phosphide by China's Ministry of Commerce in February 2025, a price decoupling between Eastern and Western indium markets has occurred. Furthermore, the purification process to achieve ultra-high purity indium of 6N (99.9999%) grade or higher, involving multi-stage techniques like electrolysis, vacuum distillation, and zone refining, further restricts the availability of raw materials for substrate manufacturing. This upstream scarcity is progressively impacting the midstream segment.
Midstream Substrate: An Oligopoly with High Barriers
The market for substrates is highly consolidated, with the transition to 6-inch wafers and lengthy expansion cycles being core challenges. While the substrate market by volume is smaller than that for epitaxial wafers, its technical barriers and market concentration are significantly higher. Substrate manufacturing is a crystallization process from scratch, with core difficulties including: 1) a severe dependency on ultra-high purity 6N+ grade indium, representing a primary bottleneck; 2) the extreme difficulty of crystal growth, requiring precise control under high temperature and high pressure, coupled with long equipment lead times and slow yield improvements; 3) a highly oligopolistic structure, with Sumitomo Electric (42%), AXT, Inc. / Beijing Tongmei (36%), and JX Nippon Mining & Metals (13%) collectively controlling over 90% of the global InP substrate market. Capacity expansions are anticipated from Sumitomo's Osaka plant and AXT's fundraising efforts. In contrast, epitaxy—growing thin films on substrates—has lower market concentration and higher shipment volumes. Consequently, InP substrates are viewed as the segment with the highest technical barriers, the tightest supply-demand dynamics, and the most critical bottleneck within the current AI optical module supply chain.
Downstream Dynamics: EML Shortages and InP's Irreplaceable Role
Shortages of Externally Modulated Lasers (EML), with deliveries reportedly pushed to 2027 or later due to capacity being allocated to companies like NVIDIA, are driving diversification in technological pathways. With an estimated 40-60% supply gap for 800G EMLs, the industry is accelerating the adoption of Continuous Wave (CW) lasers combined with Silicon Photonics. However, the InP base for CW lasers still relies on suppliers like Sumitomo and AXT, meaning demand for InP is increasing, not decreasing. The industry is evolving through generations from EML (current) to CW+SiPh (rapid development) to CPO with External Light Sources (2026-2028), and towards monolithic III-V on silicon (2030+).
Profit Concentration and Investment Implications
Industry profits are becoming concentrated in two areas: substrate/epitaxy and high-end optical chips. Vertically integrated manufacturers and leading substrate producers are gaining amplified pricing power. Substrates face the highest technical and concentration barriers. The epitaxy segment is more fragmented due to commercially available MOCVD tools, but involves high customization and strong customer loyalty. Downstream EML/CW/UHP chips and CPO/ELS modules show potential for 2-3x increases in average selling prices. Integrated Device Manufacturers (IDMs) with full-stack capabilities (e.g., Lumentum Holdings Inc, Coherent Corp) and leading substrate material companies (e.g., Sumitomo, AXT/Tongmei) are positioned to benefit from scarcity premiums.
The industry is characterized by supply shortages and robust demand. Investors may consider monitoring leading companies in the optical communication space, such as AXT, Inc. (AXTI), Sumitomo Electric Industries (5802.T), IQE plc (IQE.L), Applied Optoelectronics, Inc. (AAOI), Lumentum Holdings Inc (LITE), Coherent Corp (COHR), and Ciena Corporation (CIEN), for potential dual improvement in profitability and valuation.
Key risks include: 1) Demand volatility due to Cloud Service Provider capital expenditure falling short of expectations; 2) Disruptions in upstream indium supply and export controls; 3) Substrate and epitaxy capacity expansion and yield ramp-up progressing slower than anticipated; 4) Technological substitution and the pace of CPO adoption being slower than expected.
Comments