Running on AWS? Here's the honest math: ECS Fargate vs Kubernetes.
We spent years building for Kubernetes before betting the company on ECS. Not because K8s is bad — because for most teams on AWS, Fargate wins on cost and ops, and almost nobody builds tooling for it. Here's the full reasoning.
Kubernetes on AWS (EKS)
AWS runs the K8s control plane. You manage worker nodes, patching, scaling, and add-ons (CNI, CSI, ingress). Pay per EC2 instance — idle capacity on every node bills 24/7. The ecosystem is massive, but the configuration surface matches it.
ECS Fargate
Lower cost, less opsAWS runs the control plane and the compute. Per-task, per-second billing — stop the task, stop paying. Fargate Spot gives ~70% discount for fault-tolerant workloads. Schedule non-prod to business hours, cut 65–77%. Zero idle — pay per task, not per node.
ECS Fargate + Fortem
FortemAll the cost and ops benefits of Fargate, plus a control plane for your entire fleet. Scheduling, cloning, cost tracking, developer self-service, AI diagnostics — on top of your existing Terraform. The tooling gap that every ECS team eventually hits, solved.
If you're on Kubernetes today
Keep it. Seriously. If your team runs K8s well, migrating to ECS to save on tooling is solving the wrong problem — the K8s ecosystem is mature and you've already paid the learning cost.
Where this page matters for you:
- —You also run ECS workloads— many AWS shops do, usually with zero tooling around them. That's exactly the gap Fortem covers.
- —You're starting a new project on AWS — the default choice deserves a second look. Per-task billing and zero node ops compound fast.
- —Your EKS bill is mostly idle dev/staging capacity — the scheduling math (65–77% off non-prod) applies to any stack that can stop; Fargate just makes stopping trivial.
We're not going to tell you K8s is dying. We're telling you ECS Fargate is underserved — and we'd rather own that side of the market.