AWS Nitro Sandboxes give every EKS pod and ECS task its own VM boundary, at no extra cost

AWS Nitro Sandboxes give every EKS pod and ECS task its own VM boundary, at no extra cost

AWS has launched Nitro Sandboxes in all commercial Regions: lightweight virtual machines run by the Nitro Hypervisor that can each host a single EKS pod, ECS task or standalone VM. The point is isolation. Every workload gets its own hardware-enforced VM boundary instead of sharing a kernel with its neighbours, so untrusted code, multi-tenant SaaS workloads, regulated data and AI agents that run generated code can share one EC2 instance without the usual container-escape worries. AWS says enabling it for EKS or ECS is a one-line change, that Pod Identity, VPC CNI, EBS, EFS, S3 and CloudWatch keep working as before, and that performance is close to native EC2. Sandboxes also support cryptographic attestation, so secrets can be released only to a sandbox that proves its integrity. Teams that want full control can run them directly on EC2 with their own AMI, orchestration and VMM, with Amazon Linux and Firecracker supported out of the box. There is no extra charge, and it works on most Nitro-based instance types. If you currently rely on Kata Containers, gVisor or one-tenant-per-node pools to get stronger isolation, it is worth testing whether this lets you consolidate nodes and simplify that setup.

This is a summary by our content curator. Read the original at AWS What's New: https://aws.amazon.com/about-aws/whats-new/2026/11/nitro-sandboxes-eks-ecs-ec2/.