InP (indium phosphide) shortage emerges as critical supply bottleneck for AI optical interconnect production

Lumentum CEO Michael Hurlston has warned that indium phosphide (InP) supply-demand gap now exceeds that of DRAM and NAND, making it the most critical constraint on AI data center optical interconnect expansion.

Despite operating multiple InP wafer fabrication facilities, Lumentum faces widening shortages as demand outpaces production. Orders from telecom customers have jumped from hundreds to hundreds of millions of devices, driven by NVIDIA and hyperscale operators.

The transition to 800G and 1.6T optical modules is rapidly increasing InP substrate and epitaxial wafer demand, but global capacity expansion has lagged. Downstream customers now place orders further in advance and pay substantial deposits; inventory across the supply chain continues to tighten, driving prices higher.

To address the shortage, major AI companies are investing directly in upstream production. In March 2026, NVIDIA committed $2 billion each to two key suppliers and signed long-term supply agreements. One supplier has already doubled its 6-inch wafer capacity and plans another expansion by the end of 2027, while the other is upgrading its 6-inch production line with mass production expected in 2028. A leading Japanese supplier has announced a multi-year expansion program, though industry analysts expect the supply imbalance to persist beyond 2027.

Indium phosphide, a III-V compound semiconductor, serves as the foundation of high-speed optical communications. Unlike silicon, InP efficiently generates light at wavelengths with minimal fiber transmission loss and offers high electron mobility, making it essential for next-generation optical transceivers.

In silicon photonics, silicon can guide, modulate, and split light but cannot generate it. Both conventional pluggable transceivers and emerging co-packaged optics (CPO) architectures therefore rely on InP-based lasers. For long-haul coherent communications, InP remains the only commercially proven substrate for high-performance electro-absorption modulated laser (EML) chips.

The shortage reflects a structural mismatch between rising demand and constrained supply. AI data centers drive higher optical interconnect density, increasing InP consumption per module as transmission speeds rise. CPO architecture adoption adds further demand for high-power laser sources.

Supply is constrained by limited raw materials, manufacturing complexity, and industry concentration. Indium is primarily recovered as a by-product of zinc refining, making output expansion difficult. China, which accounts for roughly 70% of global refined indium production, has tightened export controls on InP and indium since 2025, further pressuring the global supply chain.

Manufacturing poses another bottleneck. Large-diameter InP substrates require highly sophisticated crystal-growth technology, with yield improvements taking years to achieve. Global production capacity remains concentrated among a few suppliers, limiting capacity expansion.

China has established a relatively complete InP ecosystem spanning raw materials, substrates, epitaxy, photonic chips, and optical modules.