Google launches ace service to provide AI agents with dedicated computing resources

Google has introduced the Agent Compute Engine (ACE), a new Google Cloud service designed to offer AI agents dedicated computing resources. The service provides each AI agent with a secure, persistent, and scalable 'personal computer' through isolated hardware microVMs. ACE is optimized for the lifecycle of AI agents, enabling rapid startup and independent kernel access with root permissions. The service also ensures security through hardware isolation with Cloud Hypervisor, preventing credential storage within sandboxes to reduce risks of misuse or data leaks. ACE's key feature is its ability to pause and resume operations automatically when agents are idle, allowing for efficient resource management. Google claims that pausing and resuming 200,000 agents takes less than 500 milliseconds, with shared memory and snapshots boosting server capacity. Initially, ACE will support Gemini, Antigravity, and internal model evaluation processes. Developers will soon be able to use ACE via Google Cloud, managing agent environments with full control over images, identities, and network access, while Google handles resource scheduling and scaling.