# Google Cloud Plans Vera Rubin Clusters With Up to 960,000 GPUs

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/37163
Published: 2026-10-21T23:31:56.000Z
Updated: 2026-10-21T23:31:56.000Z
Section: Technology

> A single site could host up to 80,000 GPUs, while the 960,000 figure covers a connected cluster spanning multiple locations. General availability is planned for November 2026.

Google Cloud plans to connect as many as 960,000 NVIDIA Vera Rubin GPUs across multiple sites, setting a maximum cluster scale for its upcoming A5X infrastructure.

The 960,000 figure is the planned ceiling for a connected cluster, not a count of chips already installed. Google Cloud’s Virgo Network can link up to 80,000 Rubin GPUs at one site; reaching the larger figure would require connecting locations.

Through A5X bare-metal instances, customers can use Rubin NVL72 systems connected by Google Cloud’s Virgo Network. Google Cloud plans general availability in November 2026 and intends to expand access into local regions across Europe, the Middle East and Africa to address data residency and sovereignty requirements. It also plans to add Google Kubernetes Engine features for managing distributed AI workloads.

Ineffable Intelligence, a London-based company, was A5X’s first customer and is applying it to large-scale reinforcement learning, in which AI models learn from feedback. The planned rollout adds a cloud option for training and running AI systems on NVIDIA’s Vera Rubin platform.