Nvidia, Tower Semiconductor Unveil 1.6T Silicon Photonics Modules for AI Networking
Nvidia and Tower Semiconductor have developed 1.6T silicon photonics-based optical modules to boost AI data center network throughput. Tower’s advanced packaging integrates photonic transceivers with Nvidia’s chips, promising multi-terabit links and reduced latency for next-generation AI infrastructure.
1. Collaboration Overview
Nvidia and Tower Semiconductor have partnered to co-develop 1.6-terabit optical modules targeting AI data center networking. The collaboration leverages Tower’s wafer-level packaging expertise to embed photonic components directly onto Nvidia’s networking chip platforms.
2. Silicon Photonics Module Specs
Each module delivers 1.6 terabits per second throughput using silicon photonics transceivers and advanced co-packaged optics. The design integrates lasers, modulators and photodetectors into a compact package optimized for high-speed data transmission.
3. Data Center Network Benefits
These optical modules enable multi-terabit links between AI accelerators, reducing signal loss and power consumption compared to traditional copper interconnects. The enhanced bandwidth and lower latency support larger-scale GPU clusters and real-time AI workloads.
4. Commercialization Timeline
Tower plans initial sampling of the 1.6T modules in Q4 2026, with volume shipments slated for early 2027. Nvidia intends to deploy the modules in its next-generation data center switches and AI networking gear.