NVIDIA’s Rubin Lands Inside Google’s Virtual Machine, Stretching Multi Site Clusters to Nearly 1 Million GPUs

Google launched A5X cloud instances for agentic AI workloads and expanded its AI infrastructure partnership with NVIDIA.

The new setup will support NVIDIA Vera Rubin GPUs and ConnectX-9 networking on Google Cloud.

Using the Virgo-based scaling platform, clusters can grow to 80,000 GPUs in one site and 960,000 across multiple sites.

Google says the move should improve token throughput and lower inference costs for large-scale AI training and deployment.