Lumentum Holdings is undershipping customer demand for electro-absorption modulated lasers (EMLs) by approximately 30%, creating a supply bottleneck for AI networking infrastructure that is expected to persist through 2027.
All EML production capacity at Lumentum is committed under long-term agreements running through the end of calendar 2027. The company increased its indium phosphide manufacturing capacity by more than 20% in the December 2025 quarter, with additional expansions planned.
The capacity additions have not closed the supply gap. As Lumentum brings new production online, demand growth continues to outpace supply increases, widening the imbalance between what customers order and what the company can deliver.
Lumentum's optical communications segment (OCS) order backlog has climbed above $400 million, with most orders scheduled for delivery in the second half of 2026. The backlog represents firm customer commitments that cannot be fulfilled within standard lead times.
EML components are critical for high-speed data transmission in AI training clusters and inference deployments. Virtually every major AI networking cluster relies on Lumentum optical products to connect GPUs and move data between servers.
The supply constraint affects hyperscale data center operators including Amazon Web Services, Microsoft Azure, and Google Cloud, which are racing to build AI infrastructure. These companies have locked in long-term supply agreements with Lumentum to secure component allocations through 2027.
For investors, the supply dynamics create a bifurcated opportunity set. Lumentum trades at a revenue multiple reflecting its sold-out production capacity and pricing power in a constrained market. Data center operators face potential delays in AI infrastructure buildouts if they lack secured component supply.
Alternative optical component suppliers including Coherent, II-VI (now Coherent), and NeoPhotonics face similar capacity constraints. The industry-wide shortage stems from the specialized nature of indium phosphide semiconductor manufacturing, which requires expensive cleanroom facilities and long capital investment cycles.
Companies with existing long-term supply agreements hold a competitive advantage in deploying AI infrastructure at scale. Those without secured component allocations may face 12-18 month lead times for new orders, delaying data center expansion plans and AI service launches.