Key points
- Google announced Agent Compute Engine, which runs each AI agent in its own microVM secured by the company's Titanium chip; preview opens in the coming months.
- Idle agents pause automatically, and shared memory between VMs lowers per-agent cost. The service already powers Gemini.
- Google says its open-sourced Agent Substrate runtime fits ten times as many agents per server, with suspend and resume under 500 milliseconds in a 200,000-agent test.
- New controls add token tracking, model routing and third-party identity support, including Ping Identity and SailPoint, across Cloud Run, GKE and managed endpoints.
Why it matters for marketers: Cheaper, isolated agent hosting and built-in spend and identity controls could lower the cost and risk of running always-on marketing agents at scale.
Google announced Agent Compute Engine, a service that places each AI agent inside its own isolated microVM. Google says the VMs are secured by its Titanium chip. The service already runs behind Gemini and will open for preview in the coming months.
Efficiency is the main pitch. When an agent goes idle, its VM pauses and releases the machine for other work, and shared memory between VMs reduces what each one costs. Google also open-sourced Agent Substrate, a Kubernetes runtime that, by the company's account, fits ten times as many agents on one server. In a test with 200,000 agents, Google says suspend and resume took under 500 milliseconds.
Google is also adding oversight tools: token tracking, model routing and support for third-party identity providers such as Ping Identity and SailPoint. These controls are set to extend to Cloud Run, GKE and managed endpoints.
For marketers and content teams, the news is mostly about economics and risk. Agents that handle campaign tasks, content workflows or customer interactions often sit idle between jobs, so pausing them could make always-on setups cheaper to run. Per-agent isolation and identity integration may also help with security reviews, especially for agencies handling several clients' data.
What to watch: the preview timing, pricing once it is available, and whether independent tests match Google's figures. Teams already building on Cloud Run or GKE should check how the new tracking and routing controls fit their current setup.
Sources
This brief is our own summary. Read the full reporting at the source.