Google Cloud schedules Nvidia Vera Rubin A5X availability for November

Google Cloud plans to make its A5X computing instances generally available in November, Nvidia said on October 21. European availability will also expand, although the announcement did not give a separate date for that rollout.

A5X combines Nvidia's Vera Rubin NVL72 systems and ConnectX-9 networking hardware with Google's Virgo network and liquid cooling. Google Cloud plans infrastructure spanning multiple locations with capacity for as many as 960,000 Rubin GPUs. That figure describes planned scale rather than an announced completed deployment.

Ineffable Intelligence, the London company founded by David Silver, received the first customer access to A5X. It is using the systems for reinforcement learning research, developing AI that improves through experience. Nvidia also said WPP tested creative AI workflows on A5X, with people approving actions before execution.

The November launch extends access beyond those early uses to customers running model training, inference and reinforcement learning. Nvidia and Google Cloud are also examining Nvidia's DSX MaxLPS technology to improve energy efficiency and support more computing within existing power limits.