Ultrafast photonics for next-generation compute

LightScale builds ultrafast photonics for AI infrastructure: optical switching at packet cadence, lower communication power at any reach, and adaptive network fabrics, with optical compute and switching for quantum computing ahead.

optical I/O · 64 in, 64 out 64 × 64 any-to-any

Photonic network switches

Scale1

Reroutes any-to-any at packet cadence, entirely in light. Built for HYPERLANE, our topology-on-demand fabric.

SXNR-001-64 · 64 × 64 Pilot
Higher port counts coming soon.

Explore Scale1

optical die-to-die · GPU to GPU two dual-die GPUs · one Photon¹ link

Optical die-to-die interconnect

Photon1

Sits beside a GPU or memory controller and carries its die-to-die traffic over light, so reticle-sized dies and their memory work as one system, targeting well under a picojoule per bit at any reach.

Explore Photon1

Front elevation of the Pod-1 ORV3 switching system: five cabinets side by side, each filled with SXNR-001-64 ORV3 switches with full-width staggered ventilation grilles and locking rack ears, with the LightScale mark on the base panel

Rack-scale photonic switching

Pod1

Full racks of Scale1 switches, connected by an all-optical fabric. Dedicated switching capacity for AI infrastructure.

PIPELINE · FWD / BWD GRADIENT ALL-REDUCE MoE ALL-TO-ALL FABRIC RESHAPES BETWEEN PHASES · PACKET CADENCE

One programming model

LUMA

A single API spans switching and compute: developers write to LUMA, not to individual devices.

Explore LUMA

The future of compute needs a new interconnect layer

Ultrafast, low-energy, programmable photonics, starting with Scale1 and pointing further, to optical compute and to switching for quantum computing.

Contact team