AWS Lambda MicroVMs: Firecracker-isolated sandboxes for running untrusted code

At its New York Summit, AWS introduced Lambda MicroVMs, a new serverless compute primitive aimed at a problem that has been awkward to solve: giving every end user or session its own isolated environment to safely run code the application developer didn't write. Think AI coding assistants, interactive notebooks, data-analytics sandboxes, vulnerability scanners, and game servers running user scripts. Until now teams had to choose between virtual machines (strong isolation, slow to start), containers (fast, but a shared kernel that needs heavy hardening for untrusted code), or functions (great for request-response, but not built to hold session state).

MicroVMs are built on Firecracker — the same lightweight virtualization that already backs Lambda at enormous scale — so each session gets true VM-level isolation with no shared kernel. The model is image-then-launch: you package a Dockerfile plus code as a zip in S3, Lambda builds and boots it once, then snapshots the running memory and disk. Every subsequent launch resumes from that snapshot instead of cold-booting, so even multi-gigabyte sessions come back near-instantly. A running MicroVM keeps memory, disk, and processes across a session, can auto-suspend during idle (you pay only for storage while paused), and supports up to 8 hours of runtime.

For an AWS-first shop this is worth a look: it removes the need to build and operate custom virtualization just to run untrusted or AI-generated code. It's available now in US East (N. Virginia, Ohio), US West (Oregon), Europe (Ireland) and Asia Pacific (Tokyo) on ARM64, with up to 16 vCPUs, 32 GB memory and 32 GB disk per MicroVM. Lambda Functions still own event-driven workloads; MicroVMs complement them for stateful, multi-tenant code execution.

This is a summary by our content curator. Read the original at AWS News Blog: https://aws.amazon.com/blogs/aws/run-isolated-sandboxes-with-full-lifecycle-control-aws-lambda-introduces-microvms/.