Lumentum is undershipping AI infrastructure customers by roughly 30% as photonics component demand outpaces production capacity. The company's entire externally modulated laser (EML) output is locked into long-term agreements running through calendar year 2027.
The optical communications segment (OCS) holds a backlog exceeding $400 million, with most units scheduled for second-half delivery. Indium phosphide manufacturing capacity jumped over 20% in the December quarter, marking one of the fastest fab expansions in the sector.
Capacity additions are not closing the supply gap. As Lumentum brings new production online, order volumes accelerate faster, widening the imbalance. The company now supplies photonics components to virtually every major AI network deployment.
Photonics manufacturers face a structural constraint: specialized semiconductor fabrication requires 18-24 month lead times for meaningful capacity expansion. Indium phosphide wafer production cannot scale at the pace of AI datacenter buildouts, creating sustained shortages through at least 2027.
Long-term supply agreements provide revenue visibility rarely seen in semiconductor components. Customers are locking capacity years in advance, accepting extended delivery schedules rather than risk component availability. This shifts photonics suppliers from cyclical to infrastructure-critical vendors.
The supply crunch extends beyond Lumentum. Coherent and other photonics specialists report similar capacity constraints in laser components, optical transceivers, and silicon photonics integration. AI training clusters require 10-100x more optical interconnect bandwidth than traditional datacenters, driving exponential component demand.
Stock implications center on multi-year revenue locks and pricing power. Photonics suppliers can prioritize margin expansion over volume growth when capacity-constrained. The $400 million backlog at Lumentum alone represents nearly 40% of trailing twelve-month OCS revenue, providing cushion against broader semiconductor volatility.
Investors should track quarterly capacity utilization rates and backlog-to-revenue ratios. Supply-demand normalization would signal margin pressure, but current trajectories suggest sustained pricing strength through 2027. AI infrastructure buildouts show no signs of slowing, with hyperscalers announcing $200+ billion in combined 2026 capex focused on compute and networking.