Marvell Demonstrates Silicon Photonics Light Engine for Low-power, Rack-scale Interconnect in AI Networks

Core Insights - Marvell Technology, Inc. is showcasing its 1.6T silicon photonics light engine at OFC 2025, designed for next-generation AI scale-up network requirements [1][4] - The 1.6T light engine supports eight channels of 200 Gbps PAM4 optical interconnect, providing a low power and highly integrated solution for mid-density rack-scale connectivity [1][3] - The product is positioned to facilitate the transition from passive copper interconnects to optimized optics in AI scale-up domains [2][4] Product Features - The 1.6T light engine is a quarter-scale version of the previously demonstrated 6.4T light engine, featuring 200G electrical and optical interfaces [3] - It consumes less than 5 picojoules per bit (pJ/bit) under typical operating conditions, consolidating hundreds of components into a single package [3][8] - The engine is designed for use in linear-drive pluggable optics (LPO) modules and can accelerate time-to-market for ecosystem partners [2][8] Market Positioning - Marvell is well-positioned to co-develop custom co-packaged optics (CPO) systems needed for high-density, row-scale applications as the industry transitions from copper to optics [4] - Demand for the 1.6T optics is strong, with linear drive pluggables offering a lower power option for customers [4] - The 1.6T light engine is currently sampling to select customers, indicating a move towards commercialization [5]