According to NVIDIA's official blog, Lambda is utilizing NVIDIA Photonics Co-Packaged Optics (CPO) switches to connect a GB300 NVL72 rack cluster comprising over 10,000 GPUs. Company testing shows that rack-level network latency is reduced by approximately 4%, pod-level latency by up to approximately 10%, while throughput remains identical to traditional pluggable optical transceiver solutions. These are specific test results disclosed by the manufacturer and customer, and do not imply that all data centers will achieve identical performance.