Google Cloud A5X (NVIDIA Vera Rubin NVL72)
Google Cloud's A5X instances on NVIDIA Vera Rubin NVL72, announced at GTC Berlin with general availability due in November 2026.
Overview
Google Cloud and NVIDIA announced A5X instances at GTC Berlin on October 21, 2026, built on Vera Rubin NVL72 systems with ConnectX-9 SuperNICs, Google's Virgo Network and fourth-generation liquid cooling. General availability is 'coming next month' (reported as November 2026), with planned expansion to Europe. Ineffable Intelligence was the first customer to access A5X, and WPP reported that workloads that normally need a multi-node cluster ran on a single A5X instance. Google plans to scale to as many as 960,000 Rubin GPUs across sites; that is a scaling target, not installed capacity. Pricing, per-instance GPU counts and regions are not published.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- Built on NVIDIA Vera Rubin NVL72 rack-scale systems
- ConnectX-9 SuperNICs and Google Virgo Network
- Fourth-generation liquid cooling
- GA expected November 2026; European expansion planned
- Target scale of up to 960,000 Rubin GPUs across multiple sites
Use cases
- Large-model training and inference on Rubin-generation GPUs
- Running multi-node workloads on a single instance (WPP early test)
Pricing
unknown; no pricing published.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 21 Oct 2026
- Record updated
- 28 Oct 2026
- Record last checked
- 28 Oct 2026