Google Cloud A5X instances powered by NVIDIA Vera Rubin are scheduled for general availability in November 2026, NVIDIA said in an October 21 report. Google Cloud also plans to expand availability to Europe and scale its Vera Rubin infrastructure to as many as 960,000 Rubin GPUs across multiple sites.

Key Takeaways

  • A5X general availability is due next month; European expansion and the deployment of up to 960,000 GPUs remain planned.
  • Ineffable Intelligence was the first A5X customer, using the platform for large-scale reinforcement learning.
  • WPP demonstrated creative agents on A5X, with some initial test workloads fitting on one instance instead of a multi-node cluster.

Table of Contents

What Google Cloud A5X brings together

A5X combines NVIDIA Vera Rubin NVL72 systems, NVIDIA ConnectX-9 SuperNICs, Google Cloud’s Virgo Network and its fourth-generation advanced liquid-cooling technology. The instances are intended to support training, inference and large-scale reinforcement learning.

The deployment brings Vera Rubin into Google Cloud’s AI Hypercomputer, integrating compute, networking and storage. NVIDIA describes the system as supporting the repeated cycle of generating experience, assessing results and updating models that reinforcement learning requires.

The report does not give a timetable for reaching the planned maximum of 960,000 GPUs or for expanding availability to Europe.

Ineffable puts reinforcement learning first

London-based Ineffable Intelligence, founded by AlphaGo creator David Silver, was the first customer to access A5X. It is using the infrastructure to develop systems that learn continuously from experience.

Silver, Ineffable’s founder and CEO, said the research requires substantial computing capacity and a close connection between experience generation and training. He pointed to connectivity between the Vera CPU and Rubin GPUs, as well as between GPUs within a rack, as enabling the company to expand its reinforcement-learning research.

WPP tests multistep creative agents

WPP demonstrated its creative-agent reference architecture on A5X using NVIDIA Nemotron models and NVIDIA OpenShell. The agents translate conversational requests into workflows including video generation, upscaling and restyling, with human approval required before execution.

OpenShell supplies governed execution sandboxes to help control access to sensitive brand assets and creative tools. In WPP’s initial tests, workloads that typically needed a multi-node cluster ran on a single A5X instance. The report does not provide broader benchmark results.

Power optimization remains exploratory

Google Cloud and NVIDIA are also exploring NVIDIA DSX MaxLPS capabilities for Vera Rubin infrastructure. The work is intended to improve performance per watt and expand capacity within existing data-center power budgets; the report does not announce a completed deployment or quantified efficiency gains.

Conclusion

The next milestone is A5X general availability in November. Early customer work demonstrates reinforcement-learning and creative-agent uses, while European expansion, the larger GPU rollout and power-optimization work remain plans or exploratory efforts.

Read More News

More FinanceVelix news

Disclaimer

This FinanceVelix article provides general information, not personalized financial, investment, legal, or tax advice. Market prices, interest rates and economic conditions can change; investments carry a risk of loss. Facts reflect the material available when published and may change. Check important information independently and consult a qualified professional when needed. Mentioning an asset, company, or service is not an endorsement.