Lightmatter has joined Nvidia’s NVLink Fusion ecosystem , delivering CPO and NPO devices optimized for Nvidia’s photonics and SerDes technology. The company’s bi‑directional optical link architecture will reduce fiber and connector usage by 50% , cutting data center cabling costs and deployment complexity.

XPU chips using Lightmatter optical interconnects can directly interface with Nvidia switch chips, breaking traditional I/O bottlenecks for large‑scale AI clusters.

CEO Nick Harris said integrating Passage CPO into NVLink Fusion unlocks frontier LLM potential. Nvidia’s Ashish Karandikar noted that Lightmatter’s photonic engines improve flexibility and power efficiency for hyperscale AI infrastructure.

Lightmatter’s Passage 3D stacked silicon photonics and Guide VLSP optical engines efficiently connect thousands to millions of processors, eliminating data transfer bottlenecks and redefining next‑gen AI data center architecture.

ICgoodFind: Lightmatter‑Nvidia collaboration accelerates CPO commercialization, cutting AI data center cost and complexity while boosting bandwidth.