Lambda's cutting-edge AI network slashes power use, boosts reliability with NVIDIA's photonic switches
Original headline: Lambda Boosts AI Factory Resiliency and Efficiency With NVIDIA Photonics — blogs.nvidia.com
More in Semiconductor and chip news
Cloud provider Lambda is deploying NVIDIA's Photonics co-packaged optics (CPO) switches to connect massive AI clusters, linking over 10,000 GPUs across NVIDIA GB300 NVL72 racks. By integrating the optical engines directly onto the switch package instead of using separate pluggable transceivers, the system aims to create faster, more resilient networking for large-scale AI model training and inference.
The upgrade delivers tangible improvements: network latency dropped by 4% at rack scale and up to 10% at pod scale, while maintaining throughput. More crucially, it achieves up to 5x better power efficiency and 15x higher reliability than conventional optics, with cooler operating temperatures (36°C vs. 51°C) and fewer failure-prone components. For engineers building AI factories, this represents a significant step toward eliminating network bottlenecks that can stall thousands of GPUs during critical training jobs.
Read the original at blogs.nvidia.com
Satirical commentary on real tech news, summarised with AI. Read the original source for the facts. How ZeroDaily works
