New See exactly what you're overpaying AWS in under 60 seconds. Estimate your AWS savings →

AWS Fargate Pricing Explained With Rates and Monthly Examples

AWS Fargate pricing is based on provisioned CPU, memory and billable running time. In US East N. Virginia, Linux x86 costs $0.04048 per vCPU-hour and $0.004445 per GB-hour. Additional configured storage, EKS cluster fees and supporting services can add charges. Running tasks remain billable while idle.
Picture of Aman Singh
Technical Content Writer, FinOps

Aman Singh

Technical Content Writer, FinOps
November 17, 2026

·

16 min read
AWS Fargate Pricing Explained With Rates and Monthly Examples
In this article
Key takeaways
1
Fargate bills provisioned resources throughout each task or pod’s billable lifetime, including idle time.
2
ECS on Fargate has no separate orchestration fee. EKS cluster fees and supporting services need separate budget lines.
3
Review resource sizes and running patterns before committing. Compute Savings Plans discount eligible CPU and memory; Spot suits interruption-tolerant Linux ECS tasks.
AWS Fargate pricing estimates often begin with one task’s hourly rate multiplied by a month. Production spending also depends on how many copies run, when they scale and which supporting services the application needs. The examples below account for those changes before assessing cost controls and commitments.

AWS Fargate pricing at a glance

Fargate runs containers without making you manage their underlying servers. ECS manages container workloads, while EKS provides managed Kubernetes. Fargate works with either. 

A vCPU is a virtual processor; RAM is working memory. The AWS regional price catalog lists these US East N. Virginia On-Demand rates, checked November 13, 2026.

AWS Fargate pricing by operating system and architecture
Configuration Per vCPU-hour Per GB-hour of memory Windows licence per vCPU-hour
Linux x86 $0.04048 $0.004445 None
Linux Arm $0.03238 $0.00356 None
Windows x86 $0.046552 $0.00511175 $0.046
Windows CPU and licence charges are added together. The Fargate pricing conditions limit Windows and Arm options to ECS. AWS defines GB here as 1,024³ bytes, equivalent to GiB.

Resources come in supported pairs. A small Linux ECS task with 0.25 vCPU and 0.5 GiB costs approximately $0.01234/hour for CPU and memory. That is a configuration example, not a flat starting fee.

Is Fargate free

Fargate has no dedicated free compute allowance. Eligible introductory AWS credits are separate account offers; their terms determine what they offset.

How Fargate billing works

An ECS task runs one or more containers together. An EKS pod is Kubernetes’ unit for running one or more containers together.

To calculate AWS Fargate pricing, count allocated resources and billable time. Under AWS’s duration rules, billing runs from image download until termination, rounded up to the next second. Each Linux task or pod run has a one-minute minimum; each Windows task run has a five-minute minimum.

For identical tasks:
CPU and memory cost = billable task-hours × [(vCPUs × CPU rate) + (GiB × memory rate)]
Ten tasks running two billable hours each create 20 task-hours. Add Windows licensing separately where applicable. Before estimating, collect the region, whether you use ECS or EKS, provisioned vCPU and memory per task or pod, task or pod counts and billable runtimes, configured ephemeral storage, and enabled supporting services. 

For ECS Linux platform 1.4.0 or later, 20 GiB of temporary storage is included; configuration can increase it to 200 GiB. Images occupy some of that space. Additional configured storage costs $0.000111/GB-hour in this region: 30 GiB for 720 hours adds 10 × 720 × $0.000111 = $0.80.

Running idle versus stopped

A quiet task still holds its assigned capacity. No incoming requests does not mean no charge. Terminating it ends that resource billing, subject to minimums; separate resources and commitments may remain payable.

ECS and EKS cost differences

The control plane coordinates workloads. ECS on Fargate has no separate orchestration fee, while EKS cluster pricing adds $0.10/hour under standard Kubernetes version support or $0.60/hour under extended support.

On EKS, application requests are only the starting point. AWS adds 256 MB for Kubernetes components and rounds capacity to a supported shape. 

Check the pod’s CapacityProvisioned annotation with kubectl describe pod <pod-name> -n <namespace> to confirm its billed CPU and memory. 

Consider one simple pod requesting 1 vCPU and 2 GiB for a 30-day month:

1 vCPU and 2 GiB requested → add 256 MB of Kubernetes memory overhead → round up to the supported 1 vCPU/3 GiB shape.
Budget line Calculation Subtotal
Pod CPU and memory 720 × ($0.04048 + 3 × $0.004445) $38.75
Standard-support cluster fee allocated in full 720 × $0.10 $72.00
Combined Before supporting services $110.75
This assigns the whole cluster fee to the example. Shared clusters need an allocation method. The cluster fee continues while the cluster exists, even between pod runs. Optional paid control-plane capacity is excluded. EKS Fargate uses Linux x86 and does not support Fargate Spot.

Additional costs beyond Fargate compute

A complete AWS Fargate pricing estimate follows the application beyond its containers:
  • Networking: Applicable internet or cross-zone transfers, public IPv4 addresses, NAT gateways or interface endpoints.
  • Load balancing: Load-balancer hours and capacity usage when enabled.
  • Observability: Log ingestion, retention and enabled monitoring features.
  • Storage: Container images in a registry and persistent files or database storage.
VPC networking charges and CloudWatch charges have their own meters. Private networking does not automatically require NAT; suitable endpoints can serve supported AWS traffic, with their own charges. Include only your actual architecture, and allocate shared services consistently.

AWS Fargate pricing examples

These illustrative estimates use Linux x86 in N. Virginia, USD On-Demand rates and 30 days. Each ECS task has 1 vCPU, 2 GiB and included storage:
Task-hour rate = $0.04048 + 2 × $0.004445 = $0.04937.
Lifetimes include image pull and termination. Figures exclude supporting services, taxes, credits and discounts. Round after calculating, not after every task.

Always on versus scheduled jobs

Operating pattern Billable task-hours CPU and memory subtotal
Two tasks continuously running 2 × 24 × 30 = 1,440 $71.09
One task running two hours daily 1 × 2 × 30 = 60 $2.96
The scheduled task terminates between runs. These are different operating patterns, not equivalent-performance alternatives.

Why baseline estimates fall short

Now take the continuously running service. Suppose it adds six tasks for two hours on 20 busy days. During 20 deployments, two additional tasks overlap for ten minutes beyond the baseline and peak counts.

ECS rolling deployments can temporarily increase task counts; these particular durations and frequencies are assumptions, not measured norms.
Component Additional or baseline task-hours CPU and memory cost
Baseline 1,440 $71.09
Peak tasks 6 × 2 × 20 = 240 $11.85
Deployment overlap 2 × (10 ÷ 60) × 20 = 6.67 $0.33
Total Approximately 1,686.67 $83.27
That is $12.18, or 17.1%, above the baseline-only estimate in this scenario. Peaks create most of the difference. These are additional task-hours, not hidden fees, and supporting services are still separate.

How to reduce Fargate costs

Start Fargate cost optimization with the quantities behind the bill. Review CPU and memory alongside latency, errors and peak demand before choosing smaller supported task sizes. Low average CPU alone does not establish safe downsizing.

ECS Service Auto Scaling changes task count under configured policies; it does not automatically resize each task. Schedule suitable non-production services or jobs, and confirm tasks actually terminate when unnecessary.

Compatible Linux ECS workloads can use Arm rates, approximately 20% below the x86 CPU/memory rates shown here. Test image compatibility and performance first. Compute savings do not automatically reduce networking or logging charges.

Fargate versus EC2 costs

EC2 can cost less when containers fit efficiently onto instances that stay usefully occupied. Fargate can avoid unused host capacity when tasks stop between jobs and removes host-management work. Compare deployable capacity, runtime, availability, storage, discounts and operational effort rather than matching vCPU labels alone.

Our EC2 versus Fargate cost comparison examines those trade-offs and break-even considerations in detail.

Fargate Spot pricing

Spot runs tasks on spare capacity. AWS advertises up to 70% below regular rates for interruption-tolerant Linux ECS tasks on x86 or Arm. Rates vary; use the current regional Spot rates rather than assuming a fixed percentage.

AWS can reclaim capacity with a two-minute interruption warning. Queue workers and restartable batch jobs are candidates when unfinished work can be recovered. Capacity is not guaranteed. Design retries and retain appropriate On-Demand capacity where availability requires it. Savings Plans do not add another discount to Spot usage.

Compute Savings Plans for Fargate

Compute Savings Plans cover eligible Fargate CPU and memory, including ECS and EKS, in exchange for a one- or three-year dollar-per-hour commitment. EC2 Instance Savings Plans and EC2 Reserved Instances do not cover Fargate.

For Linux x86 in N. Virginia, matching CPU/memory rates in the AWS Savings Plans catalog produce these reductions, checked November 13, 2026:
Term and payment Reduction versus On-Demand resource rates
One year No Upfront 20%
Three years No Upfront 45%
Three years All Upfront 52%
These are native resource-rate comparisons, not guaranteed total-bill savings. Windows licensing, additional storage, EKS cluster fees and supporting services remain separate.

For AWS Fargate pricing estimates, review existing coverage after sizing changes. Size commitments against eligible hourly usage you expect to sustain. Value the hourly commitment at Savings Plans rates, rather than On-Demand-equivalent spending. 

Other eligible EC2 or Lambda usage can consume the same plan; uncovered usage remains On-Demand. The commitment remains payable when matching usage falls.

Also read our Lambda and Fargate Savings Plans coverage guide for mixed-service eligibility and sizing.

How Usage.ai manages Fargate commitments

Once task sizes and running patterns are sensible, the next question is how much steady Fargate spend your existing Savings Plans leave uncovered. At Usage.ai, our read-only Savings Test reviews billing, usage and existing coverage. Purchases require separate authorization and permissions.

With our Flex Insured Commitments, teams can get up to 52% savings on Fargate CPU and memory through a three-year All Upfront AWS Compute Savings Plan with none of the commitment risk.

If an eligible Flex Commitment costs more than equivalent pay-as-you-go usage, we provide Cashback protection for the qualifying difference. Cashback is paid 90 days after it accrues. Our protection does not cancel the underlying commitment, and we do not automatically protect existing customer-owned commitments as Flex Commitments.

We charge an agreed percentage of realized savings, so the fee applies when verified savings are generated.

Build your Fargate budget

Build AWS Fargate pricing estimates in AWS Pricing Calculator from valid shapes, billable task-hours, cluster fees and supporting costs. Before committing, compare with actual charges by service and usage type in Cost Explorer.

To explore commitment coverage for your eligible Fargate baseline, book a demo to discuss a read-only Savings Test.
Fargate commitment review
Review your eligible Fargate baseline

Discuss existing coverage, potential commitment savings and the billing data needed for a personalized assessment.

Frequently asked questions

How much does Fargate cost monthly?

AWS Fargate pricing has no fixed monthly fee. One Linux x86 task with 1 vCPU and 2 GiB costs about $35.55 for 720 billable hours in N. Virginia, before supporting services. Task count, resources, runtime and applicable cluster charges determine the wider budget.

Does Fargate have a free tier?

There is no dedicated recurring Fargate compute allowance. Eligible introductory AWS credits may offset qualifying charges under your account’s terms. Included temporary storage is part of the service configuration, not free CPU or memory.

Does Fargate charge while idle?

Yes. Allocated resources remain billable while a task or pod runs, even without requests. Termination ends that resource billing subject to minimums. Separate resources and existing Savings Plans obligations can continue costing money.

Is Fargate cheaper than EC2?

Neither is always cheaper. AWS Fargate pricing follows requested resources and task duration; EC2 costs depend on instances, task packing and supporting resources. Compare equivalent availability and operational requirements, then include applicable discounts and management effort.

Do Compute Savings Plans cover Fargate?

Yes, eligible CPU and memory for ECS and EKS can receive Compute Savings Plans rates when hourly commitment remains available. Windows OS licensing and additional storage are not discounted. EC2 Instance Savings Plans and EC2 Reserved Instances do not provide Fargate coverage.

Share
Facebook
X
LinkedIn
Reddit
See your savings first
More from Usage.ai
Latest from our blogs