Google Cloud and NVIDIA Take Vera Rubin to General Availability With A5X Instances

Google Cloud and NVIDIA said A5X instances accelerated by NVIDIA Vera Rubin NVL72 will reach general availability next month, with a planned expansion of availability into Europe. The two companies framed the infrastructure around training, inference and large-scale reinforcement learning for cloud customers.

The A5X configuration brings together Vera Rubin NVL72 systems, NVIDIA ConnectX-9 SuperNICs, Google Cloud's Virgo network and the cloud provider's fourth-generation advanced liquid-cooling technology. The stated target is demanding agentic inference and reinforcement learning workloads at scale.

Scale is the headline number: Google Cloud plans to grow its Vera Rubin-based footprint to as many as 960,000 Rubin GPUs across multiple sites.

Ineffable Intelligence, a London-based company founded by AlphaGo creator David Silver, was the first customer to access A5X. Silver said his team is building superlearners that improve through their own experience, which demands substantial compute and a tight connection between generating experience and training on it.

WPP has also demonstrated a creative agent reference architecture on A5X instances using NVIDIA Nemotron models and NVIDIA OpenShell, part of the reference architecture for safe AI agents. The agents turn conversational requests into multistep creative workflows such as video generation, upscaling and restyling, with human approval before execution. In initial tests, workloads that typically required a multi-node cluster ran on a single A5X instance.

Security is part of the pitch. Google Cloud will be among the first providers to offer NVIDIA Confidential Computing at rack scale on bare-metal instances such as A4X Max machines powered by GB300 NVL72, and it is making Confidential Computing generally available on G4 VMs with RTX PRO 6000 Blackwell Server Edition GPUs. Customers including Vertebrae, Polaris Cloud and Confidential Core AI already use Confidential G4 for conversational privacy, verifiable agent execution and policy-governed enterprise intelligence.

On efficiency, the two companies said they are exploring NVIDIA DSX MaxLPS capabilities for Vera Rubin-based infrastructure to improve performance per watt and fit more capacity inside existing data-centre power budgets.

Keeping experience generation and training inside one tightly connected system is what makes continuous learning practical at scale. The near-term tests are whether general availability lands on schedule next month, how quickly European capacity follows, and whether the 960,000-GPU plan materialises as described.