NVIDIA said in a blog post on 21 October that Google Cloud will make its new A5X instances generally available next month. An instance is a rented cloud server. A5X is built on NVIDIA's Vera Rubin NVL72 rack systems, plus Google's own network and liquid cooling. NVIDIA said Google Cloud plans to grow this setup to as many as 960,000 Rubin GPUs across several sites. London firm Ineffable Intelligence was the first customer to use A5X. NVIDIA also said Google Cloud is making Confidential Computing generally available on its G4 virtual machines.
NVIDIA said London marketing group WPP ran its creative AI agents on A5X. In WPP's initial tests, work that usually needs a multi-node cluster ran on one A5X instance. The post gives no prices, no launch day and no list of European regions. It does not say when the 960,000 GPU figure would be reached. NVIDIA said it and Google Cloud are still exploring a power-saving feature called DSX MaxLPS. No wider performance tests were included.
Teams that plan AI training or agent work on Google Cloud will have a new hardware option next month. Prices and European locations are the details buyers need to compare costs and decide where data can sit.