AI Demand Sparks Global Capacity Race for Indium Phosphide Optical Interconnects
Semiconductor giants and Chinese firms are aggressively expanding production to close a 70% supply gap in critical InP materials.
The explosive growth of AI data centers is triggering a global "capacity arms race" for Indium Phosphide (InP) optical interconnects. As AI clusters scale, the industry is shifting from electrical to optical interconnects to resolve critical bottlenecks in data movement speed and power efficiency.
To address a supply-demand gap of nearly 70% that is expected to persist through 2028, major players are committing billions in capital expenditures. Japan's JX Advanced Metals plans to invest up to ¥120 billion to expand its InP substrate capacity to up to 10 times current levels by FY2030. In China, DSBJ's subsidiary, Source Photonics, is investing US$1.2 billion into a high-speed optical module and optical chip expansion project in Changzhou. Additionally, Sanan Optoelectronics has moved to mass-produce 6-inch InP optical chips, aiming to increase its current capacity of 2,750 wafers per month to nearly 6,000.
The Shift to Optical Architectures
This surge in investment comes as the industry evolves through several architectural stages: traditional pluggable optics, Near-Packaged Optics (NPO), and Co-Packaged Optics (CPO). While CPO provides the highest efficiency, it remains plagued by maintenance and yield challenges. This has resulted in a multi-route coexistence of laser sources, including VCSEL, MicroLED, and Silicon Photonics combined with Continuous Wave (CW) lasers, all of which rely heavily on the underlying semiconductor value chain.
Strategic Value Chain Reshaping
The transition to optical interconnects is fundamentally reshaping the semiconductor ecosystem, creating a deep dependency on InP materials and specialized foundry capacity. Currently, international giants control approximately 95% of the global InP substrate market. For Chinese firms, this "arms race" is not merely about capacity but is a strategic effort to localize the supply chain and break an existing monopoly. This shift extends benefits beyond chip designers to MOCVD equipment manufacturers, raw material suppliers of gallium and germanium, and OSATs specializing in heterogeneous bonding.
Future Outlook
Industry observers are now watching whether these massive capital injections can close the supply gap fast enough to keep pace with AI hardware roadmaps. While the trend toward localization and capacity expansion is clear, the long-term dominance of specific architectures—particularly the viability of CPO versus NPO—will determine which specific InP production technologies become the gold standard for the next generation of AI infrastructure.