Google unveils Agent Compute Engine to run each Gemini agent in hardware‑isolated microVMs
Google announced Agent Compute Engine, a service that runs each AI agent inside its own hardware‑isolated microVM secured by Google's Titanium chip and says it already powers Gemini. The service — which underpins the open‑sourced Agent Substrate Kubernetes runtime — pauses idle VMs to free resources, uses shared memory to cut per‑agent costs, and (Google says) can fit up to ten times as many agents per server; it will open for preview in coming months and adds token tracking, model routing, and third‑party identity provider support extending to Cloud Run, GKE and managed endpoints.