Lambda, a cloud provider, is deploying NVIDIA Photonics co-packaged optics (CPO) switches to strengthen the networking layer of its AI infrastructure. According to the input, the switches are being used to connect NVIDIA GB300 NVL72 racks spanning more than 10,000 GPUs, with the aim of delivering a network that is fast, efficient and resilient.

Why it matters

Training large AI models requires thousands of GPUs to operate together, which places heavy demands on the interconnecting network. Addressing the network component with co-packaged optics is positioned as a way to improve both efficiency and resiliency at this scale.

Who should care

Cloud providers and organizations building or operating large-scale GPU clusters for AI training are the primary audience for this deployment.