Lambda Deploys NVIDIA Photonics to Optimize Massive AI Infrastructure

Executive Briefing
“Lambda is integrating NVIDIA co-packaged optics switches to connect more than 10,000 GPUs,...”
Key Takeaways
- 1.Cloud provider Lambda is deploying NVIDIA Photonics co-packaged optics switches to connect NVIDIA GB300 NVL72 racks featuring over 10,000 GPUs.
- 2.In tests, the technology reduced network latency by about 4% at rack scale and up to roughly 10% at pod scale while maintaining conventional throughput.
- 3.The new architecture places optical engines directly on the switch package, delivering up to 5x better power efficiency and higher reliability than conventional pluggable optics.
Lambda is integrating NVIDIA co-packaged optics switches to connect more than 10,000 GPUs, boosting reliability and power efficiency for large-scale artificial intelligence data centers.
Key Developments and Technical Context
Cloud provider Lambda is deploying NVIDIA Photonics co-packaged optics switches to connect NVIDIA GB300 NVL72 racks featuring over 10,000 GPUs.
In tests, the technology reduced network latency by about 4% at rack scale and up to roughly 10% at pod scale while maintaining conventional throughput.
The new architecture places optical engines directly on the switch package, delivering up to 5x better power efficiency and higher reliability than conventional pluggable optics.
Strategic Analysis and Industry Implications
Moving electrical-to-optical conversion closer to the switch chip shortens the signal path from inches to millimeters, eliminating retimer chips and reducing failure-prone components.
The liquid cooling solution maintains a median optics temperature of 36 degrees Celsius while allowing technicians to replace external light sources in the field when needed.
Outlook and Key Factors to Watch
The long-term operational efficiency of the system under continuous large-scale language model training workloads remains unquantified outside of initial test data.







