Marvell: 1.6T Silicon Photonics Light Engine

As AI domains scale from rack-scale to row-scale, Marvell’s 1.6T engine provides hyperscalers with a flexible, future-ready optical platform that works today and evolves toward tomorrow’s co-packaged optics (CPO) architectures.
Feb. 12, 2026

As workloads such as generative AI and high-performance computing increase east-west traffic inside data centers, legacy copper interconnects can no longer provide the reach, efficiency, or signal integrity required. Marvell’s 1.6T light engine directly solves these challenges by delivering eight lanes of 200 Gbps PAM4 optical connectivity in a single-package, ultra-low-power device optimized for both linear-drive pluggable optics (LPO) and on-board optical integration. Designed to help ecosystem partners reduce costs and speed time-to-market, the light engine consolidates hundreds of discrete components—including modulators, photodetectors, linear drivers, TIAs, firmware, and passive elements—into a ready-to-use optical engine. This dramatically simplifies module design, eliminates complex optical tuning steps, and reduces bill of materials (BOM) overhead. With a validated reference design and deployment-ready integration path, module vendors and hyperscalers can bring 1.6T LPO products to market significantly faster while lowering engineering investment.

Judge’s comment: “The high degree of device and multi-channel integration enabled in Marvell's 1.6T Silicon Photonics Light Engine will enable significant simplification for 1.6T LPO modules, while delivering important reductions in power/bit.”