NVIDIA, Google Cloud Unveil Vera Rubin-Powered A5X GPU Clusters Scaling to 960K Units

NVIDIA and Google Cloud expanded their AI partnership on April 22, 2026, launching A5X bare-metal instances with NVIDIA Vera Rubin GPUs that scale to 960,000 across sites and 80,000 in a single site, claiming 10x lower inference costs and 10x higher token throughput per megawatt. This follows Google Cloud's earlier deal granting Thinking Machines Labs expanded access to A4X Max instances using NVIDIA Blackwell architecture. The announcements highlight surging enterprise demand for NVIDIA accelerators, with customers like OpenAI and CrowdStrike already deploying GB300/GB200 systems, bolstering bullish NVDA outlook amid NVDA trading at 199.97 (+0.05%).