Optical Connectivity Within the Rack
High-performance photonic solutions delivering ultra-dense, low-latency optical links for AI cluster scale-up.

Overview
Within the rack, AI and machine learning clusters require ultra-high bandwidth density and predictable, low-latency performance to maintain synchronization and memory coherency across tightly coupled accelerators and high-bandwidth memory. Copper interconnects are reaching their physical limits in both reach and energy efficiency, restricting connectivity to within a single rack. Lumentum overcomes these barriers with advanced EML and CW laser platforms, Ultra High Power (UHP) lasers, and External Laser Source Pluggable (ELSFP) modules that extend optical connectivity across multiple adjacent racks. In co-packaged optics (CPO) architectures, ELSFP modules centralize light generation for improved thermal management and scalability, while EML- and CW-based solutions deliver power-efficient, short-reach optical links. Together, they enable high-performance, rack-level interconnects that scale efficiently for next-generation AI systems.
