Google Cloud is getting ready to rent out Nvidia's newest AI racks at huge scale. Its A5X instances, built on Nvidia Vera Rubin NVL72 systems, will reach general availability next month (November), with expansion to Europe planned, Nvidia said in a blog post on Wednesday, Oct. 21, 2026, timed to its GTC Berlin conference. Google Cloud plans to scale its Vera Rubin-based infrastructure to as many as 960,000 Rubin GPUs across multiple sites. That is a stated plan, not installed capacity.

What A5X is

Each A5X system combines Vera Rubin NVL72 with Nvidia ConnectX-9 SuperNICs, Google Cloud's Virgo Network and Google's fourth-generation liquid cooling, and plugs into Google Cloud's AI Hypercomputer as one integrated compute, networking and storage platform, Nvidia said. It is aimed at agentic inference and large-scale reinforcement learning, where a model repeatedly generates experience, scores the results and updates itself.

First customers

  • Ineffable Intelligence, the London company founded by AlphaGo creator David Silver, was the first customer to access A5X. "We're building superlearners that continually improve through their own experience," Silver said in Nvidia's post. WebProNews notes Ineffable raised a $1.1 billion seed round in April.
  • WPP, the London-based marketing group, demonstrated creative AI agents on A5X using Nvidia Nemotron models and Nvidia OpenShell, with human approval before anything runs. In WPP's initial tests, workloads that typically needed a multi-node cluster ran on a single A5X instance, Nvidia said.

Confidential computing and power

Nvidia said Google Cloud will be among the first cloud providers to offer Nvidia Confidential Computing at rack scale on bare-metal instances such as A4X Max (GB300 NVL72), and is making Confidential Computing generally available on G4 VMs with RTX PRO 6000 Blackwell Server Edition GPUs. The two companies are also exploring Nvidia's DSX MaxLPS to squeeze more capacity out of fixed data-center power budgets. Nvidia CEO Jensen Huang said in his GTC Berlin keynote that Vera Rubin is in full production, WebProNews reported, citing Nvidia's live keynote updates.

What it means for Tesla

The same Vera Rubin chips are at the heart of Musk-company AI plans: on SpaceX's Q3 call on Tuesday, Musk said SpaceX plans a training cluster of 440,000 Nvidia Vera Rubin chips by the end of next year's second quarter (SpaceX Q3 story). Google's 960,000-GPU plan shows how hard the big clouds are competing for the same supply. Also from GTC Berlin: Nvidia's OpenShell agent-safety runtime and 14-day AI factory pods.