In brief
- General availability for A5X instances is planned for next month, with expanded availability in Europe also under consideration.
- The system combines NVIDIA Vera Rubin NVL72, ConnectX-9 SuperNICs and Google Cloud Virgo Network components.
- NVIDIA says Google Cloud plans to scale its Vera Rubin infrastructure to up to 960,000 Rubin GPUs across multiple facilities.
- Ineffable Intelligence is the first customer to access A5X, while WPP has demonstrated creative agent workflows on the platform.
NVIDIA announced on 21 October 2026 that Google Cloud plans to make its Vera Rubin-based A5X cloud instances generally available next month. The company also aims to expand availability in Europe. According to NVIDIA, the infrastructure is designed for compute-intensive workloads such as AI training, inference and large-scale reinforcement learning.
What components make up the A5X infrastructure?
A5X combines the NVIDIA Vera Rubin NVL72 system with Google Cloud networking and cooling components. The announcement lists NVIDIA ConnectX-9 SuperNICs, Google Cloud Virgo Network and the company’s fourth-generation advanced liquid-cooling technology. NVIDIA says these components are integrated to support large-scale workloads that run experience generation and model training together.
Google Cloud plans to scale its Vera Rubin-based infrastructure to up to 960,000 Rubin GPUs across multiple facilities. This figure represents the company’s stated scaling target, not its current capacity. Although general availability for A5X instances is planned for next month, the announcement does not specify pricing or an exact date for availability in Europe.
What are the first users doing with A5X?
London-based Ineffable Intelligence was named the first customer to access A5X. Founded by David Silver, creator of AlphaGo, the company is using the platform for large-scale reinforcement learning research aimed at developing systems that learn continuously from experience. This work links the steps of generating experiences, evaluating outcomes and updating models.
London-based marketing and communications company WPP also showcased a reference architecture for creative agents on NVIDIA Vera Rubin-based A5X. According to NVIDIA, these agents can turn natural-language requests into multistep tasks such as video generation, image upscaling and restyling. Human approval is required before the processes proceed. OpenShell helps manage access to sensitive brand assets and creative tools through controlled environments in which the agents operate. WPP’s initial tests reportedly ran workloads that would normally require a multi-node cluster on a single A5X instance.
What work is underway on privacy and energy use?
The announcement also covers other NVIDIA infrastructure on Google Cloud alongside Vera Rubin A5X. Google Cloud is described as one of the first cloud providers to offer rack-scale NVIDIA Confidential Computing on bare-metal instances such as A4X Max, which uses NVIDIA GB300 NVL72. Confidential Computing is reported to be generally available on G4 virtual machines equipped with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.
On energy efficiency, Google Cloud and NVIDIA are exploring NVIDIA DSX MaxLPS features for Vera Rubin-based infrastructure. The aim is to help improve performance per watt and increase capacity within existing data center power budgets. The announcement presents this as work in progress, not as a completed product feature.
What this means for architecture and visualization workflows
The A5X announcement does not specify support for any particular architectural visualization software, render engine or design application. For architecture firms and 3D artists, the takeaway is therefore not that a specific application will run faster on A5X, but that infrastructure intended to scale compute-intensive AI workloads in the cloud is being planned. NVIDIA’s announcement includes no measurements of architectural rendering performance or application compatibility.
Teams evaluating AI agents in visual production workflows may also need to consider human approval and controlled access to sensitive assets, as in WPP’s example. Firms exploring cloud use will need to verify the general availability timeline, announced pricing and compatibility with the tools they use. The announcement provides no information about availability or licensing conditions specific to Turkey.
Sources
1 sourceSource texts are not republished; short quotes are marked, everything else is our own summary and commentary.
For architecture firms and visualization teams, A5X’s practical value lies not in being presented as a direct render accelerator today, but in its goal of scaling compute-intensive AI workloads in the cloud. The WPP example shows that human approval and controlled access to brand assets may also be part of creative workflows; however, it does not prove that architectural production software is supported.
For teams in Turkey, it is important to wait for general availability and the announcement of pricing and licensing conditions before making a decision. No details on hardware or software compatibility have been provided, so the impact on existing workflows can only be assessed once concrete support information and test results are available.
Frequently asked questions
When will Google Cloud A5X instances become generally available?
NVIDIA announced that general availability is planned for next month. Expanded availability in Europe is planned, but no exact date has been specified.
Which NVIDIA components are used in the A5X infrastructure?
A5X combines NVIDIA Vera Rubin NVL72 and ConnectX-9 SuperNICs with Google Cloud Virgo Network. The announcement also mentions Google Cloud’s fourth-generation advanced liquid-cooling technology.
Does A5X accelerate architectural visualization software?
The announcement provides no measurements for architectural visualization software, rendering performance or application compatibility. The infrastructure targets AI training, inference and large-scale reinforcement learning workloads.
