Short answer: for a small team running a handful of services, ECS is usually cheaper and simpler. EKS starts paying off when you need the Kubernetes ecosystem, portability, or a platform team already fluent in it.
What you pay for in each
| Cost item | Amazon ECS | Amazon EKS |
|---|---|---|
| Control plane | No charge | US$ 0.10 per cluster per hour (standard support) |
| Compute | EC2 or Fargate at normal prices | EC2 or Fargate at normal prices |
| Versions and upgrades | Managed by AWS | Kubernetes versions to upgrade; extended support costs more per hour |
| Add-ons | Mostly native integrations | Ingress controller, autoscaler, observability: you choose and maintain |
One EKS cluster in standard support (730 hours × US$ 0.10), before any node or Fargate task. Separate clusters per environment multiply it.
The hidden cost: operations
The control plane fee is small next to the time spent operating Kubernetes: version upgrades, add-on compatibility, RBAC, networking choices and on-call knowledge. For a team of three, that time is usually the deciding factor, not the US$ 73.
When to choose ECS
- A few to a few dozen services, mostly HTTP APIs and workers.
- The team wants AWS-native integrations (ALB, IAM roles per task, CloudWatch) with little platform work.
- No requirement to run the same platform outside AWS.
When to choose EKS
- You depend on Kubernetes tools: Helm charts, operators, service mesh, GitOps.
- Portability across clouds or on-premises is a real requirement, not a hypothetical one.
- There is a platform team, or budget for one, to own upgrades and add-ons.
Fargate is not automatically cheaper. For steady, always-on workloads, EC2 with Savings Plans is often cheaper; Fargate wins on low or bursty usage and on not managing nodes.
Next in this hub: a 30-day measured comparison of the same service on ECS and EKS, with the full bill.