Google's Agent Compute Engine gives every AI agent its own isolated microVM

Google has introduced Agent Compute Engine, a service that gives every AI agent its own isolated virtual machine. The company says the service already powers Gemini and will open for preview in the coming months.

Each agent runs inside a microVM secured by Google's Titanium chip. When an agent goes idle, its virtual machine pauses automatically and frees the hardware for other work, and memory shared between VMs lowers the cost of running each additional agent.

Efficiency is the heart of the release. Virtual machines are among the largest cost factors in operating agentic systems, so pausing idle agents and sharing memory goes straight at the two biggest line items in an agent fleet.

The engine also underpins Agent Substrate, an open-sourced Kubernetes runtime that Google says can fit ten times as many agents on a single server. In a test with 200,000 agents, suspend and resume took less than 500 milliseconds, according to the company.

On oversight, Google is adding token tracking, model routing and support for third-party identity providers including Ping Identity and SailPoint, with those controls extending to Cloud Run, GKE and managed endpoints.

Read together, the release treats isolation, cost and governance as one problem. An agent that holds credentials and takes actions needs a boundary around it, an agent that idles most of the day should not hold hardware, and an enterprise needs to know which agent spent which tokens under whose identity.

What to watch next is the preview date and pricing, whether Agent Substrate attracts contributors outside Google, and whether rival clouds respond with comparable per-agent isolation. The competitive ground here is infrastructure for running agents, not model quality.