NVIDIA and Google Cloud Launch Vera Rubin Infrastructure for Agentic AI Workloads
Google Cloud and NVIDIA announce general availability of A5X instances next month, featuring Vera Rubin GPUs and advanced liquid cooling for large-scale reinforcement learning.
Key points
- General availability of NVIDIA Vera Rubin-based A5X instances arrives next month on Google Cloud.
- London-based Ineffable Intelligence is the first customer, using the platform for continuous reinforcement learning.
- Google Cloud plans to scale the infrastructure to 960,000 Rubin GPUs across multiple sites.
Google Cloud and NVIDIA announced the general availability of NVIDIA Vera Rubin infrastructure, with instances launching next month. The release introduces an integrated platform for training, inference, and large-scale reinforcement learning to cloud customers, marking a significant step in deploying agentic computing systems at scale.
The new A5X instances combine NVIDIA Vera Rubin NVL72 systems with NVIDIA ConnectX-9 SuperNICs. They also integrate Google Cloud’s Virgo Network and fourth-generation advanced liquid-cooling technology. This hardware stack is designed to support demanding agentic inference and reinforcement learning workloads that require high bandwidth and thermal efficiency.
Infrastructure Scale and Early Adopters
Google Cloud intends to expand its Vera Rubin-based infrastructure to as many as 960,000 Rubin GPUs across multiple sites. This expansion follows earlier reporting by Threat Times on October 21, 2026, regarding the traction of NVIDIA Vera CPUs among European sovereign AI partners and the introduction of the DSX POD to accelerate deployment.
Ineffable Intelligence, a London-based technology firm founded by AlphaGo creator David Silver, was the first customer to access A5X instances. The company is using the platform to develop “superlearner” systems that continuously improve through experience. David Silver stated that the platform’s compute performance and high-bandwidth connectivity allow them to push the scale of their reinforcement learning research further.
Enterprise Applications and Security
WPP, a London-based marketing and communications company, demonstrated its creative agent reference architecture on A5X instances. Using NVIDIA Nemotron models and NVIDIA OpenShell, the agents convert conversational requests into multistep creative workflows such as video generation and upscaling. WPP noted that workloads previously requiring multi-node clusters now run on a single A5X instance.
To address security concerns for proprietary models and sensitive data, Google Cloud is offering NVIDIA Confidential Computing at rack scale. This is available on bare metal A4X Max instances powered by NVIDIA GB300 NVL72. Additionally, Confidential Computing is generally available on G4 VMs with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, supporting customers like Vertebrae and Polaris Cloud in maintaining conversational privacy and verifiable agent execution.
What should you do?
Security and technology professionals should monitor vendor announcements for specific deployment dates of Vera Rubin instances. Evaluate whether current infrastructure supports the bandwidth requirements of large-scale reinforcement learning workloads. Review data protection policies for agentic systems, particularly when using third-party cloud infrastructure for sensitive model training. Stay informed about confidential computing options to safeguard proprietary data during inference and fine-tuning processes.
What to do and how to stay safe: NVIDIA
- Verify that your organization’s data governance policies address the use of third-party infrastructure for training proprietary AI models.
- Assess the network bandwidth requirements of your reinforcement learning workloads to ensure compatibility with high-speed interconnects like ConnectX-9.
- Monitor vendor updates regarding the geographic expansion of Vera Rubin availability, particularly for European data sovereignty requirements.
- Investigate confidential computing capabilities offered by cloud providers to protect sensitive data during inference and fine-tuning phases.
Step-by-step guide: Mobile Device Security: How to Protect Data Beyond Passwords
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
When will NVIDIA Vera Rubin instances be generally available on Google Cloud?
General availability of the instances is scheduled for next month, according to the NVIDIA blog post.
Which company was the first to use the new A5X instances?
Ineffable Intelligence, a London-based technology company founded by David Silver, was the first customer to access A5X.
How many Rubin GPUs does Google Cloud plan to deploy?
Google Cloud plans to scale its Vera Rubin-based infrastructure to as many as 960,000 Rubin GPUs across multiple sites.
