Early tests by advertising group WPP showed that AI agent workloads utilizing NVIDIA Nemotron and OpenShell, which previously required a multi-node cluster, can run on a single A5X instance. Google is also making Confidential Computing generally available on G4 VMs with NVIDIA RTX PRO 6000 Blackwell GPUs, with future plans for A4X Max instances powered by GB300 NVL72. To manage power constraints, the companies are deploying NVIDIA DSX MaxLPS technology to increase deployment density while maintaining existing data center power limits.
Sunday, October 25, 2026
Story
Google Cloud Scales to 960,000 Nvidia GPUs to Cut AI Clusters to Single Node
AI4 days ago
Ineffable Intelligence, the venture founded by AlphaGo creator David Silver, has become the first customer to access NVIDIA Vera Rubin NVL72 systems on Google Cloud's new A5X instances. These instances will become generally available to all customers next month and are slated for expansion into Europe. Google will scale this infrastructure across multiple data center sites to support up to 960,000 Rubin GPUs, using ConnectX-9 SuperNICs, Virgo networking, and fourth-generation liquid cooling to optimize reinforcement learning and large-scale AI training.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.