Marvell Technology has unveiled a 1.6T silicon photonics light engine designed for low-power, rack-scale optical interconnects in AI networks. Demonstrated at OFC 2025 in a 1.6T OSFP linear pluggable optics (LPO) module, the integrated optical engine supports 200Gbps per lane across eight channels using PAM4 modulation. The device includes a silicon photonics chip, linear driver, transimpedance amplifier (TIA), embedded microcontroller, and firmware—all consolidated into a single package to streamline system integration and reduce latency.
Targeted at AI server deployments that are transitioning from passive copper to optical interconnects, the light engine provides an energy-efficient alternative to copper for moderate-density rack-scale and future row-scale networks. It consumes under 5 picojoules per bit, including laser power, and enables longer reach at significantly lower power and latency. The 1.6T light engine builds on the company’s 6.4T light engine for co-packaged optics (CPO), introduced in 2024, and is designed to meet the bandwidth and integration requirements of emerging 1.6T interconnects while simplifying the commercialization path for Marvell’s ecosystem partners.
Marvell positions the 1.6T light engine as a foundation for both pluggable and on-board optics, as well as next-generation CPO systems. Sampling is underway to select customers, with the aim of enabling faster deployment of 200G/lane AI fabrics in hyperscale environments.
Key Points:
• Marvell demonstrates 1.6T silicon photonics light engine in LPO module at OFC 2025.
• Supports 8×200Gbps PAM4 channels with under 5 pJ/bit power consumption.
• Integrates silicon photonics chip, linear driver, TIA, and control firmware in one package.
• Provides a low-power, longer-reach alternative to copper in rack-scale AI infrastructure.
• Serves as a design base for future co-packaged optics (CPO) deployments.
“Marvell is advancing power-efficient AI scale up with its 1.6T light engine, delivering deployment-ready solutions for module vendors and hyperscalers,” said Josef Berger, associate vice president of marketing, Cloud Platforms at Marvell.
