Google Cloud and NVIDIA announced the upcoming general availability of NVIDIA Vera Rubin infrastructure, introducing A5X instances designed for large-scale agentic inference and reinforcement learning workloads. The platform integrates NVIDIA Vera Rubin NVL72, ConnectX-9 SuperNICs, and Google Cloud’s liquid-cooling technology. Google Cloud plans to scale this infrastructure to up to 960,000 Rubin GPUs across multiple locations, with plans for European availability.

Ineffable Intelligence, a technology firm founded by David Silver, was named the first customer using A5X to advance continuous-learning ‘superlearner’ systems. Marketing firm WPP also demonstrated creative agent workflows using NVIDIA Nemotron models on the platform. Additionally, Google Cloud announced rack-scale NVIDIA Confidential Computing on bare-metal A4X Max instances and general availability of Confidential G4 VMs to protect proprietary models and data.

Why it matters

  • Allows AI infrastructure operators to scale complex reinforcement learning pipelines using tightly integrated compute and networking.

  • Provides enterprise founders with rack-scale confidential computing options to execute proprietary agents without exposing sensitive data.

Source: blogs.nvidia.com