[NVIDIA Announces Progress on Vera Rubin Platform: CoreWeave Tests Show 10x Token Throughput Improvement per Unit Power Consumption] On July 21, NVIDIA announced that its Vera Rubin platform, designed for 'gigascale AI factories,' is expanding deployment globally. CoreWeave, Google Cloud, Microsoft Azure, and Oracle Cloud have begun adopting the Vera Rubin NVL72 system. NVIDIA stated that CoreWeave tests indicate that Vera Rubin NVL72 achieves a 10x token throughput improvement per megawatt compared to Grace Blackwell NVL72. The platform integrates Vera CPU, Rubin GPU, NVLink 6, ConnectX-9 SuperNIC, BlueField-4 DPUs, and Spectrum-6 switching systems, aiming to enhance AI training and inference efficiency. NVIDIA revealed that Vera Rubin now covers over 350 factory nodes across 30 countries, with the goal of reducing AI infrastructure costs and improving energy utilization efficiency. (Jiemian)