Google Cloud and Nvidia Announce Vera Rubin A5X Instances, With Availability Next Month
The new instances pair Vera Rubin NVL72 racks with Google's Virgo network and could scale to as many as 960,000 Rubin GPUs.
Photo: Lambtron / CC BY-SA 4.0 via Wikimedia Commons
Google Cloud and Nvidia announced on October 21 the arrival of Nvidia Vera Rubin infrastructure on Google Cloud, according to a post on Nvidia's blog. General availability of the new A5X instances is planned for next month, with a planned expansion to Europe.
The platform
- A5X instances combine Nvidia Vera Rubin NVL72 systems, ConnectX-9 SuperNICs, Google Cloud's Virgo Network and Google's fourth-generation liquid cooling.
- They are integrated into Google Cloud's AI Hypercomputer, which brings together compute, networking and storage.
- The planned scale is as many as 960,000 Rubin GPUs across multiple sites.
Early customers
London-based Ineffable Intelligence, founded by David Silver, was the first customer to access A5X and uses it for large-scale reinforcement learning. Silver said, "We're building superlearners that continually improve through their own experience."
Marketing group WPP demonstrated a creative-agent reference architecture on A5X using Nvidia Nemotron models and the OpenShell toolkit. The agents handle video generation, upscaling and restyling, with human approval before execution. In initial tests, workloads that typically needed a multi-node cluster ran on a single A5X instance, according to the post.
Confidential computing and efficiency
- Google Cloud is among the first providers to offer rack-scale Nvidia Confidential Computing on bare-metal instances such as A4X Max, which uses GB300 NVL72 systems.
- Confidential Computing is generally available on G4 virtual machines with RTX PRO 6000 Blackwell Server Edition GPUs. Vertebrae, Polaris Cloud and Confidential Core AI are early customers.
- The two companies are exploring Nvidia's DSX MaxLPS for Vera Rubin infrastructure to improve performance per watt and add capacity within existing data-center power budgets.
Context
The announcement came the same day Nvidia detailed supplier expansions to ease capacity limits. The post does not list specific Google Cloud regions or pricing for A5X.
