Lambda deploys NVIDIA Photonics co-packaged optics switches across large GPU clusters
The cloud provider says the optical switches connect more than 10,000 GPUs, with lower power draw, higher reliability and cooler-running optics than pluggable alternatives.
In plain words
Lambda, a cloud provider, is using NVIDIA's new co-packaged optics switches to link the racks of GPUs that train large AI models. These switches send data as light over fiber and place the optical parts on the switch itself, which the companies say cuts power use and reduces failures. In Lambda's tests, data moved slightly faster and the optics ran cooler than conventional pluggable modules.
What was announced
Lambda is using NVIDIA Photonics co-packaged optics switches to connect NVIDIA GB300 NVL72 racks in a deployment exceeding 10,000 GPUs, according to the NVIDIA Blog. The switches have 144 ports running at 800Gb/s, and liquid cooling carries heat away from the switch chip and the optical engines that convert electrical signals into light for fiber links. Dave Ward, Lambda's president of cloud services, said networking must keep pace with increasingly large AI clusters and described the deployment as a step toward next-generation AI infrastructure.
In Lambda's tests, network latency dropped by about 4% at rack scale and up to about 10% at pod scale, while throughput matched conventional pluggable optics across rack, scalable unit and pod configurations. NVIDIA claims the switches deliver up to 5x better power efficiency and up to 15x higher reliability than pluggable optics, using 4x fewer lasers. Lambda reports a median optics temperature of 36°C versus 51°C for pluggable optics, and each switch has 18 removable external light sources that technicians can replace in the field. The reported latency, temperature and reliability figures come from the companies themselves and have not been independently verified.
What it does not show
- All performance, efficiency and reliability figures are vendor-reported by Lambda or NVIDIA; no independent testing is cited in the source.
- The source is a vendor blog post, so deployment details such as timeline, location and cost are not covered.
For security and IT teams
- Scope
- The deployment connects NVIDIA GB300 NVL72 racks with more than 10,000 GPUs using 144-port switches running at 800Gb/s.
Sources
Source-support review completed 17 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.
Updates and corrections
No corrections so far.
First published 17 Nov 2026
New to the subject? AI from the Ground Up starts from the beginning.