AWS Lambda MicroVMs vs AWS Nitro Sandboxes
Choose AWS Lambda MicroVMs for AWS-native agent sandboxes and interactive sessions that need VM isolation, an inbound endpoint and cheap idle time via suspend; choose AWS Nitro Sandboxes for AWS teams that want a hardware-enforced VM boundary around each agent session on EKS, ECS or their own EC2 fleet, at no charge beyond the instances. AWS Lambda MicroVMs uses Firecracker microVM isolation; AWS Nitro Sandboxes uses Nitro hypervisor VM isolation.
Key differences between AWS Lambda MicroVMs and AWS Nitro Sandboxes
- Isolation: AWS Lambda MicroVMs uses Firecracker microVM; AWS Nitro Sandboxes uses Nitro hypervisor VM.
- Snapshots: AWS Lambda MicroVMs documents snapshots; AWS Nitro Sandboxes does not.
- Pause and resume: AWS Lambda MicroVMs documents pause and resume; AWS Nitro Sandboxes does not.
- Session length: 8 h (AWS Lambda MicroVMs) versus No limit; tied to parent instance (AWS Nitro Sandboxes).
- SDKs and interfaces: Only AWS Lambda MicroVMs lists Python, JavaScript/TypeScript, PHP and REST API. Only AWS Nitro Sandboxes lists Kubernetes RuntimeClass, ECS task definition and Firecracker API.
What they have in common
- Both are from Amazon Web Services.
- Both are cloud or model provider services.
- Both use microVM isolation.
- Neither is open source.
- Both list CLI.
AWS Lambda MicroVMs vs AWS Nitro Sandboxes at a glance
| Attribute | AWS Lambda MicroVMs | AWS Nitro Sandboxes |
|---|---|---|
| What it is | A Lambda resource type that runs a full Amazon Linux 2023 environment in a Firecracker microVM, restored from a pre-built snapshot, with suspend/resume and an HTTPS endpoint per VM. | An Amazon EC2 feature that runs each EKS pod, ECS task or Firecracker workload in its own lightweight VM beside your instance, with the Nitro hypervisor enforcing the boundary. |
| Type | Cloud or model provider service | Cloud or model provider service |
| Isolation | Firecracker microVM per MicroVM ("VM-level isolation with full operating system capabilities"); each runs Amazon Linux 2023 and is restored from a Firecracker snapshot of the built image | Each sandbox is a lightweight VM with its own kernel that runs on the Nitro Hypervisor beside the parent EC2 instance and is launched through Firecracker; the Nitro System isolates sandboxes from each other, and in confidential mode it also removes sandbox memory from the parent instance |
| Open source | No | No |
| Self-host / BYOC | No / No | No / No |
| SDKs | Python, JavaScript/TypeScript, PHP, CLI, REST API | Kubernetes RuntimeClass, ECS task definition, Firecracker API, CLI |
| Cold start (vendor claim) | Docs claim "near-instant launch and resume" from snapshot; the Claude Managed Agents integration page says "Resume from Firecracker snapshot in sub-second to single-digit seconds" | Not documented |
| Max session | maximumDurationInSeconds is 1-28,800 seconds (8 hours) including suspended time; the quota page lists "Maximum execution duration per MicroVM: 8 hours", not adjustable | No session limit is documented: a sandbox runs until it powers off, its init process exits or its Firecracker process stops, and every sandbox stops when the parent instance stops, reboots or terminates |
| Persistence | Suspend preserves memory and disk state (checkpointed), resume restores it; auto-suspend after a configurable idle period and auto-resume when traffic arrives; no compute charges while suspended; images are versioned Firecracker snapshots built from a Dockerfile | No snapshot and restore, pause and resume or live migration. State lives on block devices backed by instance store or EBS volumes of the parent instance, or on S3 and EFS over the network; on EKS, pods can mount EBS and EFS volumes through the CSI drivers |
| Snapshots / pause | Yes / Yes | No / No |
| GPU | Not documented — sizes are CPU/memory only (0.25-4 vCPU baseline, burstable to 4x); GPUs are not mentioned in the docs | No — GPU passthrough into a Nitro Sandbox is currently unavailable; on accelerated instances a sandbox runs CPU workloads only |
| Networking | Inbound via a per-VM HTTPS endpoint requiring a JWE auth token (HTTP/1.1, HTTP/2, WebSockets, gRPC, SSE; default port 8080); NO_INGRESS connector disables inbound; outbound is public internet by default or a customer VPC egress connector; PrivateLink supported | Each sandbox runs its own network stack. On EKS, pods keep a VPC IP through the VPC CNI, and an annotation boots a pod with no network device at all; on ECS each task gets an elastic network interface (awsvpc mode). With Firecracker you wire a TAP device yourself and must block the parent's instance metadata service (169.254.169.254), which a routed sandbox can otherwise reach |
| Pricing | US East (N. Virginia), Arm: $0.0000276944 per vCPU-second and $0.0000036667 per GB-second while running; snapshot write $0.0038/GB (suspend), snapshot read $0.00155/GB (resume or launch); image and suspended-state storage $0.08 per GB-month (one-week minimum for images); billed per second at the baseline rate plus active use above baseline | No additional charge: sandboxes run on the vCPUs and memory of the parent EC2 instance, so you pay standard EC2 rates, plus the usual EKS Auto Mode or ECS Managed Instances management fee if you use them |
| Free tier | Not documented | Not documented |
| Compliance | Not documented | Not documented |
| Regions | 14 AWS Regions as of 22 Oct 2026: launched in US East (N. Virginia, Ohio), US West (Oregon), Europe (Ireland), Asia Pacific (Tokyo); later Mumbai, Singapore, Sydney, Frankfurt, Stockholm, Hong Kong, Malaysia, Thailand and Spain | All commercial AWS Regions, on most Nitro-based instance types with Intel, AMD or Graviton processors; flex instances are excluded, as are the smallest and bare-metal sizes of most families |
| Company headquarters | Seattle, Washington, United States | Seattle, Washington, United States |
| Good fit for | AWS-native agent sandboxes and interactive sessions that need VM isolation, an inbound endpoint and cheap idle time via suspend | AWS teams that want a hardware-enforced VM boundary around each agent session on EKS, ECS or their own EC2 fleet, at no charge beyond the instances |
Verified November 14, 2026 from each provider’s documentation. Cold-start figures are vendor claims, not our measurements.
Which should you choose?
Reasons to choose AWS Lambda MicroVMs
- You want Firecracker-isolated sandboxes inside your own AWS account with IAM, VPC egress and PrivateLink
- You need sessions that sit idle for long periods without paying for compute (suspend/resume with state intact)
- You run inbound-facing workloads (dev servers, WebSockets, gRPC) that need a per-sandbox authenticated endpoint
- You are building self-hosted sandboxes for Claude Managed Agents on AWS
Reasons to choose AWS Nitro Sandboxes
- You already run agents on EKS, ECS or EC2 and want VM isolation per pod, task or session without a new vendor
- You want a boundary enforced by the Nitro hypervisor rather than a shared kernel or nested virtualisation
- You need to keep workload memory from your own node operators and gate secrets on attestation
- You want no charge beyond the EC2 instances you already pay for
Trade-offs to check first
- AWS Lambda MicroVMs: Hard 8-hour maximum lifetime per MicroVM, including suspended time
- AWS Lambda MicroVMs: GPU sizes are not documented; largest baseline is 8 GB / 4 vCPU (32 GB / 16 vCPU at peak) with 32 GB disk
- AWS Nitro Sandboxes: An EC2 feature rather than a sandbox API: no SDK for creating sessions, running commands or moving files is documented
- AWS Nitro Sandboxes: No snapshots, pause and resume, live migration or GPU passthrough
Both are worth piloting with your own image and region. See the full AWS Lambda MicroVMs profile and AWS Nitro Sandboxes profile, or the complete comparison.
More comparisons
Frequently asked questions
Is AWS Lambda MicroVMs or AWS Nitro Sandboxes better for AI agents?
What is the difference between AWS Lambda MicroVMs and AWS Nitro Sandboxes?
Which is open source, AWS Lambda MicroVMs or AWS Nitro Sandboxes?
How do AWS Lambda MicroVMs and AWS Nitro Sandboxes isolate code?
How does AWS Lambda MicroVMs pricing compare with AWS Nitro Sandboxes?
Do AWS Lambda MicroVMs and AWS Nitro Sandboxes support GPUs?
Can you self-host AWS Lambda MicroVMs or AWS Nitro Sandboxes?
How long can sessions run on AWS Lambda MicroVMs and AWS Nitro Sandboxes?
Do AWS Lambda MicroVMs and AWS Nitro Sandboxes support snapshots or pause and resume?
Sources
- docs.aws.amazon.com/lambda/latest/dg/lambda-microvms-guide.html
- docs.aws.amazon.com/lambda/latest/dg/microvms-how-it-works.html
- docs.aws.amazon.com/lambda/latest/dg/microvms-launching.html
- docs.aws.amazon.com/lambda/latest/dg/microvms-images.html
- docs.aws.amazon.com/AWSEC2/latest/UserGuide/amazon-ec2-nitro-sandboxes.html
- docs.aws.amazon.com/AWSEC2/latest/UserGuide/nitro-sandboxes-run-firecracker.html
- docs.aws.amazon.com/eks/latest/userguide/sandboxing.html
- aws.amazon.com/blogs/compute/nitro-sandboxes-hardware-isolation-for-every-eks-pod-ecs-task-and-ec2-workload/
Check before you rely on it. Products change quickly, and this page reflects public documentation as of November 14, 2026. Confirm anything that matters in the provider’s own documentation.
Cite this page
Agent Sandboxes. (2026). AWS Lambda MicroVMs vs AWS Nitro Sandboxes. agentsandboxes.org. Last reviewed November 14, 2026. https://agentsandboxes.org/compare/aws-lambda-microvms-vs-aws-nitro-sandboxes/