Introducing Scale1

SXNR-001-64 chip optical I/O · 64 in, 64 out 64 × 64 any-to-any
Top-down view of the fabric in
ORV3 · front Full-width ventilation grille
ORV3 · rear Ventilation grille 32 fibers + 2 QSFP per side · 48 V DC Ventilation grille
1U dev kit · front 64 hot-pluggable O-O ports QSFP ×4 · USB-C ×2
1U dev kit · rear PSU 1 PSU 2 1 + 1 hot-swap PSUs · IEC inlets Ventilation grille front-to-back airflow
64 × 64 Pilot

Our first photonic network switch

Scale1 is LightScale's first photonic network switch, built to turn the platform into a concrete interconnect product for AI infrastructure.

It is the first switch built for Hyperlane, our topology-on-demand network architecture, with a focus on faster data movement, lower communication overhead, and more adaptive connectivity across bandwidth-heavy workloads.

Switching Non-thermal capacitive photonic switching matrix with packet-cadence reconfiguration
Scale1 performance High radix, GHz switching rate, nanosecond switching period
Initial focus AI interconnect workloads where topology and communication efficiency directly shape system throughput
Scaling 128×128, 256×256, 384×384, 512×512, and 768×768 models coming soon
Decode throughput COMPUTE EXCHANGES WAITING 0 10 ms 20 ms 30 ms Packet fabric, median 49.5 tok/s Packet fabric, 99th pct 39.7 tok/s Hyperlane on Scale¹ 55.5 tok/s · no waiting +12% TOKENS PER SECOND AT THE MEDIAN · +40% AT THE 99TH PERCENTILE 120 cross-node exchanges per token. Direct circuits remove the electronic hops and the waiting.
GPUs per rail group · by model 0 2,048 4,096 6,144 GPUS SXNR-001-64 512 GPUs · PILOT SXNR-001-128 1,024 GPUs SXNR-001-256 2,048 GPUs SXNR-001-384 3,072 GPUs SXNR-001-512 4,096 GPUs SXNR-001-768 6,144 GPUs One switch per rail, eight rails, eight GPUs per node. The switch count stays at eight. The radix sets the GPU count.

Latency reduction

The switching path is designed to reconfigure at packet cadence, fast enough to change topology within a collective operation, not just between runs. The goal is to make connectivity changes cheap enough to use inside the data path, not as a slow control-plane action layered on top of it.

Hyperlane · topology on demand

Topology, routing, and link assignment are exposed to the control plane. Operators can match fabric shape to workload patterns (training, inference serving, MoE routing) and feed congestion and link-health signals back into reconfiguration decisions in real time.

Non-thermal switching

Scale1 uses a non-thermal photonic switching matrix, integrated at the device level rather than assembled from discrete optical components. It is designed to scale switch radix without the thermal load or per-port power overhead of conventional reconfigurable optics.

Explore LUMA

Scale1 models
Model Port count Status
SXNR-001-64 64 × 64 Pilot
  • Scale1 optical engine + ASIC
  • 96-core ARM CPU
  • 256 GB DDR5 ECC RDIMM
  • 1 TB NVMe SSD
  • 2 × NVIDIA ConnectX-7 200/400G NIC
  • 64 optical ports
  • 4 × QSFP ports
  • Open Rack V3, 48 V DC · 1U dev kit for lab
SXNR-001-128 128 × 128 Coming soon
SXNR-001-256 256 × 256 Coming soon
SXNR-001-384 384 × 384 Coming soon
SXNR-001-512 512 × 512 Coming soon
SXNR-001-768 768 × 768 Coming soon