Google Cloud will make its Nvidia Vera Rubin-based A5X instances generally available next month, with European expansion planned. The systems target large-scale model training, agent inference and reinforcement learning.
Each A5X configuration combines Vera Rubin NVL72 compute, ConnectX-9 networking, Google’s Virgo network and liquid cooling inside the company’s AI Hypercomputer platform. Google says it plans to scale Rubin infrastructure to as many as 960,000 GPUs across multiple sites, though that figure describes planned capacity rather than hardware available at launch.
Ineffable Intelligence, founded by AlphaGo creator David Silver, was the first customer to access A5X and is using it for reinforcement-learning systems. WPP also tested a creative-agent workflow that turns requests into video generation, upscaling and restyling with human approval; Nvidia says work that previously needed several nodes ran on one instance.
The announcement does not provide pricing or a detailed regional launch schedule. It also describes confidential-computing options on other Google Cloud GPU systems, while rack-scale confidential computing for Vera Rubin remains an area the companies are developing.