Google Cloud launches Vera Rubin next month, but Europe has no date

Google Cloud will make its new A5X instances, built on NVIDIA's Vera Rubin platform, generally available next month. NVIDIA said so in a blog post from GTC Berlin on Wednesday 21 October. Europe is part of the plan, but without a date: the post only says availability is planned to expand to Europe.

The headline of NVIDIA's post says the companies are announcing general availability. The text itself is more careful. It says general availability "is coming next month". For Norwegian organisations, that means this is still a plan, not capacity they can order in a European region today.

What Google Cloud is building

A5X combines NVIDIA Vera Rubin NVL72 systems with ConnectX-9 SuperNICs, Google Cloud's Virgo network and what Google calls its fourth-generation liquid cooling. The instances are aimed at what NVIDIA describes as demanding agentic inference and large-scale reinforcement learning.

According to the post, Google Cloud plans to scale its Vera Rubin infrastructure to as many as 960,000 Rubin GPUs across multiple sites. The post does not say where the sites are, how many will be in Europe, or when they will come online.

A5X becomes part of Google Cloud's AI Hypercomputer, where compute, networking and storage are sold as one integrated system. The post does not give pricing.

First customers: a research lab and an advertising group

The first customer to get access is Ineffable Intelligence. The London-based company was founded by David Silver, the creator of AlphaGo. Ineffable uses the platform for reinforcement learning aimed at systems that learn continuously from their own experience. Silver says in the post that Vera Rubin's combination of compute and bandwidth between processors and GPUs lets the company push the scale of its research further.

Marketing and communications group WPP has demonstrated a reference architecture for creative agents on A5X. The agents turn conversational requests into multistep workflows such as video generation, upscaling and restyling. A human must approve before the agents execute tasks. WPP uses NVIDIA Nemotron models and OpenShell, which is meant to provide governed sandboxes for the agents and limit access to brand assets and creative tools.

NVIDIA says workloads that typically required a multi-node cluster ran on a single A5X instance in WPP's initial tests. The post does not say which workloads, how large they were, or what it means for cost.

Confidential computing is available now

The part of the announcement that applies today is about data security. Google Cloud is making confidential computing generally available on G4 VMs with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. Confidential computing is meant to protect models and data while they are being processed, even on hardware the customer does not own. NVIDIA names Vertebrae, Polaris Cloud and Confidential Core AI as customers.

Google Cloud will also be among the first cloud providers to offer NVIDIA confidential computing at rack scale on bare-metal instances, such as A4X Max with GB300 NVL72. No date is given for this.

Google Cloud and NVIDIA are also exploring NVIDIA DSX MaxLPS for the Vera Rubin infrastructure, to get more performance per watt within data-centre power budgets. This is described as exploration, not a decision.

What leaders should note

• Vera Rubin on Google Cloud is announced for next month, but European availability has no date. Organisations that must keep data in the EU or EEA should not plan around this capacity until Google Cloud names regions.
• Confidential computing on G4 is the only part that is generally available now. It is most relevant for organisations that want to run their own models or fine-tuning on sensitive data in a public cloud. The post does not say which regions it covers.
• The WPP example shows a model where agents need human approval before they act. It is a useful reference for internal requirements, but the performance claims are NVIDIA's own and come without detail.

Sources and media

• NVIDIA: Google Cloud and NVIDIA Announce General Availability of NVIDIA Vera Rubin Infrastructure, blog post published 21 October 2026 at 12:45 Oslo time
• The source was checked on 21 October 2026.
• Illustration: AI-generated with Grok Bot / hogby.ai