Google Cloud to Deploy Up to 960,000 NVIDIA Vera Rubin GPUs

Google Cloud announced plans to deploy up to 960,000 NVIDIA Vera Rubin GPUs across its infrastructure and launched the A5X instance based on the Rubin NVL72 rack-scale system. The new A5X instance integrates NVIDIA ConnectX-9 SuperNICs alongside Google Cloud's proprietary Virgo networking and fourth-generation liquid cooling technology. This massive infrastructure commitment marks a major leap in cloud computing capabilities and AI hardware scaling, designed to support heavy agentic AI training and inference workloads. By adopting next-generation architecture at a massive scale, Google Cloud is positioning itself at the forefront of the hyper-scale AI infrastructure market. The A5X instance featuring the Rubin NVL72 system will launch next month, making Google Cloud one of the first cloud providers to offer rack-scale confidential computing on bare-metal instances. Additionally, confidential computing will be introduced on G4 virtual machines powered by NVIDIA RTX PRO 6000 Blackwell server GPUs.