Google Cloud and NVIDIA Announce Vera Rubin Infrastructure
According to NVIDIA. New integrated platform brings Vera Rubin NVL72 to Google Cloud for large-scale AI training and reinforcement learning.
- • General availability of A5X instances featuring NVIDIA Vera Rubin NVL72 technology.
- • Google Cloud plans to scale infrastructure to 960,000 Rubin GPUs across multiple sites.
- • Introduction of rack-scale Confidential Computing for secure AI agent execution.
A5X Instance Launch
Google Cloud and NVIDIA have announced the general availability of A5X instances, which integrate NVIDIA Vera Rubin NVL72, ConnectX-9 SuperNICs, and Google Cloud’s advanced liquid-cooling technology. This platform is designed to support demanding agentic inference and large-scale reinforcement learning workloads within Google Cloud’s AI Hypercomputer.
Customer Adoption and Use Cases
Ineffable Intelligence, founded by David Silver, is utilizing the platform to advance reinforcement learning systems that learn continuously from experience. WPP has also demonstrated creative agent architectures on A5X instances, enabling multistep creative workflows such as video generation and upscaling while maintaining governed access to sensitive brand assets.
Security and Sustainability
Google Cloud is expanding its Confidential Computing offerings to include rack-scale support on bare metal instances to protect proprietary models and sensitive data. Furthermore, the companies are exploring NVIDIA DSX MaxLPS capabilities to optimize performance per watt, aiming to expand capacity within existing data-center power budgets.
- 1. blogs.nvidia.comOpen source