Amazon Fargate & ECS Task Pricing: vCPU-Seconds, Ephemeral Storage, and Fargate vs EC2 Launch Type (2026)
Quick summary: Fargate bills per vCPU-second and GB-second, not per task hour. In us-east-1 a 0.25 vCPU / 0.5 GB task runs ~$9/mo; a 4 vCPU / 16 GB task ~$145/mo. Prefer arm64 + C9g/C8g when moving steady-state to ECS on EC2.
Key Takeaways
- In us-east-1 a 0
- 25 vCPU / 0
- 5 GB task runs ~$9/mo; a 4 vCPU / 16 GB task ~$145/mo
- Prefer arm64 + C9g/C8g when moving steady-state to ECS on EC2
- astro'; As of June 2026, Fargate prices Linux x86 at $0
Table of Contents
As of June 2026, Fargate prices Linux x86 at $0.04048/vCPU-hour and $0.004445/GB-hour in us-east-1; Graviton (arm64) runs ~20% lower. ECS itself has no per-task control-plane fee on EC2 launch type — you pay EC2 (or Fargate) plus optional Service Connect, CloudWatch, and data transfer.
Quantified pattern
A B2B API platform (~$22k/mo containers, 40 Fargate services) averaged 38% CPU utilization on rightsized tasks. Moving steady-state services to ECS on EC2 with Spot capacity providers cut container compute $6,800/mo (~31%) with two weeks of platform engineering. Bursty webhook workers stayed on Fargate. For net-new ECS-on-EC2 capacity providers, prefer arm64 + C9g/C8g over legacy c7g — see ECS C9g capacity-provider diagram.
Fargate vs ECS on EC2
| Factor | Fargate | ECS on EC2 |
|---|---|---|
| Ops burden | Lowest | AMI, patching, scaling |
| Billing unit | vCPU + GB seconds | EC2 instance hours |
| Best fit | Spiky / small teams | Sustained utilization |
| Spot | Fargate Spot (interruptions) | EC2 Spot via capacity providers |
Model all three paths: Fargate calculator, ECS on EC2 calculator, Fargate vs EC2 calculator.
New — June 18, 2026: ECS service auto scaling now supports 20-second high-resolution metrics (vs the default 60s). AWS measured scale-out trigger time dropping 363s → 86s (76% faster) and total scale-and-provision 386s → 109s (72% faster) — on Fargate, ECS Managed Instances, and EC2. Faster reaction lets you hold a thinner steady-state baseline and let scaling absorb spikes, trimming the always-on vCPU/GB bill (high-resolution metrics bill at standard CloudWatch rates).
If you only do one thing this week
Pull ECS CPUUtilization and MemoryUtilization for your top five services by cost. Any service under 30% average CPU and memory for 14 days is a Fargate rightsizing candidate — drop one vCPU or 1 GB and re-measure before changing launch type. For spiky services, also enable 20-second high-resolution metrics on the target-tracking policy so the thinner baseline scales out before users notice.
More in This Series
Part of the AWS Service Pricing series (June 2026).
- Previous: Part 25
- Related: Part 19 — Amazon EKS Pricing
- Browse all: Cost Optimization & FinOps category
- Hub: AWS cost optimization resources
AWS Cloud Architect & AI Expert
AWS-certified cloud architect and AI expert with deep expertise in cloud migrations, cost optimization, and generative AI on AWS.
