DigitalOcean MicroVMs: the compute foundation for building your AI agent infrastructure (private preview)
DigitalOcean has launched MicroVMs, a new compute service based on Firecracker technology that provides hardware-level isolation for container images. The service introduces automatic pausing and stateful checkpointing, allowing workloads to resume from memory and file states after idle periods.
Verified State Diff
Impact & Verification Analysis
AI infrastructure engineers, platform developers, and DevOps teams building ephemeral, agentic, or untrusted code execution environments.
This release allows developers to offload the complex 'undifferentiated heavy lifting' of virtualization and state management to DigitalOcean, significantly reducing infrastructure costs for bursty workloads while improving security posture for AI agent execution.
Full Fact Overview
DigitalOcean is leveraging Firecracker, an open-source VMM (Virtual Machine Monitor) originally developed by AWS, to offer a serverless-style compute primitive. By moving away from shared-kernel container isolation to hardware-virtualized boundaries, DigitalOcean addresses the security risks inherent in executing AI-generated or untrusted code. The technical innovation lies in the 'checkpointing' capability, which mitigates cold-start latency by allowing developers to boot, warm up, and snapshot an environment, then instantiate new VMs from that pre-warmed state. This architecture is specifically optimized for event-driven, bursty workloads like AI agents, CI/CD runners, and ephemeral sandboxes, effectively shifting the billing model from 'always-on' to 'compute-on-demand' with state persistence.