NVIDIA and Google Cloud expand the stack for agentic and physical AI

NVIDIA and Google Cloud are deepening their AI infrastructure partnership and moving it beyond generic GPU capacity toward a fuller stack for agentic and physical AI. The goal is to make it easier to build, run, and secure autonomous workflows, robots, and digital twins on Google Cloud.

What is new

In the official announcement, NVIDIA highlights several concrete updates: new Vera Rubin-based A5X bare-metal instances, a preview of Gemini on Google Distributed Cloud with Blackwell and Blackwell Ultra GPUs, confidential VMs with Blackwell, and support for NVIDIA Nemotron models and the NeMo framework inside Gemini Enterprise Agent Platform.

It is a deeply technical announcement, but the core point is simple. Google and NVIDIA are trying to make the same platform relevant for frontier models, open models, agent frameworks, and sensitive production environments. That is where many enterprises are now stuck: not on model choice alone, but on how model access, security, governance, and infrastructure fit together.

Why this matters

For executives, this is mainly an infrastructure and control story. The combination of confidential execution, distributed Gemini deployments, and support for open Nemotron models points to a future where companies can run more advanced AI closer to their own data and under tighter security and compliance constraints. It also shows how quickly hyperscalers and chip vendors are moving to own the full chain from model to production.

Source and date validation

The original source is NVIDIA’s own blog post, “NVIDIA and Google Cloud Collaborate to Advance Agentic and Physical AI,” published on April 22, 2026 at 12:00:42 UTC. That official timestamp keeps the story inside the 48-hour freshness window.

Source: https://blogs.nvidia.com/blog/google-cloud-agentic-physical-ai-factories/