Updated 21 Oct 2026 • 7 mins read
Amazon EKS Pricing Explained (2026)
Amazon EKS pricing is a $0.10-per-hour control plane, a 6x extended-support rate once a Kubernetes version ages out, and then the compute: EC2 nodes, Fargate pods, or Auto Mode with its management fee. This breakdown gives every rate in US dollars, the charges around the cluster, and three worked examples.
Amazon EKS pricing is easy to quote and hard to budget, because the number everyone quotes, $0.10 an hour for the control plane, is the smallest line on the bill. An EKS cluster also bills for the nodes its pods run on, or the Fargate capacity each pod reserves, or the management fee Auto Mode adds, and for a control-plane rate that jumps sixfold by calendar when a Kubernetes version ages out of support, plus the volumes, load balancers, gateways, addresses, and transfer that the workloads consume from the rest of AWS. The control plane is the entry fee. This guide explains everything that comes after it.
It covers every pricing component as of October 2026 in US East, the three ways to run compute and what each costs, the extended-support rule and its date, the charges around the cluster that most estimates omit, and three worked examples. For how to reduce the result, our EKS cost optimization guide picks up where this one ends. Rates are from the official EKS pricing page and the Fargate pricing page.
The short answer: Amazon EKS charges $0.10 per cluster per hour, about $73 a month, for the control plane on a Kubernetes version in standard support, and $0.60 per hour, about $438 a month, once the version enters extended support 14 months after release. Compute is billed separately in one of three ways: EC2 nodes at their instance rates (an m6i.large is $0.096 per hour, about $70 a month, in US East); Fargate pods at $0.04048 per vCPU-hour and $0.004445 per GB-hour on x86, 20 percent less on Arm; or EKS Auto Mode at the EC2 rate plus a management fee of roughly 12 percent. Around the cluster, EBS volumes, load balancers with their public IPv4 addresses, NAT Gateways, data transfer, and CloudWatch typically add 20 to 40 percent. A small three-node cluster runs about $300 a month all in; a 30-node production cluster with networking and storage can run $6,000 or more.
Key Takeaways
- The control plane is $73 a month per cluster in standard support and $438 in extended support, which switches on automatically 14 months after a version's release.
- Compute is the bill: EC2 nodes at instance rates, Fargate per vCPU and GB allocated, or Auto Mode at EC2 plus about 12 percent.
- Fargate costs more per unit than well-packed nodes and less than poorly packed ones; the crossover is around 80 percent node utilization.
- EBS, load balancers, NAT, public IPv4, transfer, and CloudWatch typically add 20 to 40 percent and are usually left out of estimates.
- The bill records nodes, not namespaces; what a team's workloads cost requires allocation from cluster metrics.
The Complete EKS Price List
US East (N. Virginia), October 2026.
| Component | Rate | Monthly (730 hours) | Notes |
|---|---|---|---|
| Control plane, standard support | $0.10 per cluster-hour | $73 | Kubernetes version within 14 months of release |
| Control plane, extended support | $0.60 per cluster-hour | $438 | Automatic from month 15 for a further 12 months; inclusive of the $0.10 |
| EC2 nodes | Instance rate: m6i.large $0.096/hr; m6i.xlarge $0.192; m7g.large (Graviton) about $0.082 | $70 / $140 / $60 per node | Spot up to ~90% off; Savings Plans up to ~66% off |
| Fargate, Linux x86 | $0.04048 per vCPU-hour + $0.004445 per GB-hour | 1 vCPU / 2 GB pod: about $36 | Per second, 1-minute minimum; 20 GB ephemeral included |
| Fargate, Linux Arm | $0.03238 per vCPU-hour + $0.003556 per GB-hour | 1 vCPU / 2 GB pod: about $29 | 20% less than x86 |
| Fargate Spot | About 70% off Fargate on-demand | 1 vCPU / 2 GB pod: about $11 | Interruptible; two-minute notice |
| EKS Auto Mode fee | About 12% of the EC2 on-demand rate, per instance-hour | m5.large: about $8.40 per node | On top of the instance; bundles Karpenter, EBS CSI, load balancer controller, node lifecycle |
| EBS (gp3) for persistent volumes | $0.08 per GB-month | 100 GB: $8 | Billed on provisioned size; persists after pods are deleted |
| Application Load Balancer | $0.0225/hr + $0.008 per LCU-hour | About $16 plus usage | One per Service of type LoadBalancer unless an ingress controller is used |
| Public IPv4 addresses | $0.005 per hour each | $3.65 each | Load balancers use one per Availability Zone |
| NAT Gateway | $0.045/hr + $0.045 per GB processed | About $33 plus usage | Image pulls and AWS API calls from private subnets pass through it |
| Data transfer | Cross-AZ $0.01 per GB each way; internet $0.09 per GB after 100 GB | Varies | Multi-AZ pods talking across zones |
| CloudWatch Container Insights and logs | Per metric and per GB ingested ($0.50/GB) | Varies | Easy to underestimate on chatty clusters |
The Three Ways to Run Compute
EC2 nodes: cheapest per unit, yours to operate
Managed node groups or self-managed nodes run on EC2 at the instance rate, so the cluster's compute cost is the node count times the instance price, and you choose families, sizes, Spot, Graviton, and commitments. It is the cheapest option per vCPU when nodes are well packed and the most work to run, since node sizing, scaling, patching, and upgrades are yours, unless you run Karpenter to automate placement.
Fargate: no nodes, billed per pod
Fargate bills each pod's allocated vCPU and memory per second, so there is nothing to pack, patch, or scale, and idle capacity costs nothing; you pay for what each pod requests while it runs. The unit price is about 20 percent above an equivalent well-packed EC2 node, so Fargate wins at low or variable utilization and loses at high, steady utilization, with the crossover around 80 percent node packing. Our Fargate pricing calculator runs the comparison for your task shape.
Auto Mode: nodes managed for a fee
Auto Mode keeps EC2 economics and removes the operations: AWS provisions and scales right-sized nodes with Karpenter, manages the storage and load balancer controllers, and patches and replaces nodes, for a management fee of about 12 percent on the EC2 rate. It suits teams without the capacity to run Karpenter and the node lifecycle themselves, and is 12 percent for convenience for teams that already do.
The Extended-Support Rule
Kubernetes releases a version roughly every four months and EKS supports each for 14 months in standard support. On the day a version passes that window, every cluster still on it moves to extended support at $0.60 per hour, six times the standard rate, automatically and without notice beyond the lifecycle calendar. Version 1.33 crossed on July 29, 2026. For a company running twenty clusters, that is $7,300 a month for not upgrading. The fix is a cadence: upgrade quarterly and the charge never applies. Our EKS cost optimization guide treats this as the first lever because it is the cheapest to pull.
Three Worked Examples
US East, always-on, three Availability Zones, on-demand rates.
| Cluster | Compute | Control plane | Around the cluster | Monthly |
|---|---|---|---|---|
| Small dev cluster: 3 x m6i.large nodes, one ALB, 100 GB EBS | $210 | $73 | ALB $16 + 3 IPv4 $11 + EBS $8 = $35 | About $320 |
| Production: 30 x m6i.xlarge, 6 ALBs, 2 NAT Gateways (8 TB), 4 TB EBS, 1 TB cross-AZ | $4,200 | $73 | ALBs and IPv4 $160 + NAT $425 + EBS $320 + transfer $20 = $925 | About $5,200 |
| Same production cluster on Kubernetes 1.33 after July 29, 2026 | $4,200 | $438 | $925 | About $5,560 |
| Fargate: 40 pods at 1 vCPU / 2 GB, one ALB | $1,440 | $73 | ALB and IPv4 $27 | About $1,540 |
The production row makes the point of this guide: the control plane is 1.4 percent of the bill, the nodes are 81 percent, and the networking and storage around the cluster, which the EKS pricing page does not mention, are 18 percent. The extended-support row shows the one line that moves by calendar rather than by usage.
What the Bill Cannot Tell You
Every line above is billed by AWS to the account. None of it is billed to a namespace, because which pods ran on which nodes is decided inside Kubernetes and never reaches billing. A team that wants to know what its workloads cost needs the node cost allocated by the capacity its pods reserved or used, plus its share of the load balancers, NAT, EBS, and transfer, which is the subject of our guides to container cost visibility and allocating Kubernetes costs. The pricing in this guide tells you what the cluster costs; those explain what each team costs.
How Opslyft Helps
Opslyft reads cluster usage metrics and allocates node cost by pod, service, and namespace, then shows it beside the EBS, load balancers, NAT, transfer, and the RDS and S3 the same workloads use, across EKS, AKS, and GKE with one model, so a team's number is its full footprint rather than its share of the nodes. That whole-bill attribution is what gave Disprz 100 percent automated allocation of 50 to 60 shared services across six platforms. For the practice that sits on top of the pricing here, start with our Kubernetes cost optimization guide.
Conclusion
Amazon EKS pricing is a small control-plane fee, a large compute bill in one of three shapes, a sixfold control-plane penalty that arrives by calendar, and a layer of networking and storage charges that estimates leave out. Budget the nodes or the pods, not the $73; decide between nodes, Fargate, and Auto Mode on utilization and operations capacity; put version upgrades on a quarterly cadence; and count the load balancers, NAT, EBS, and transfer, which are usually a fifth of the bill. Then allocate it, because the bill will never say which team spent it. For the levers that bring the number down, see our EKS cost optimization guide; for the general Kubernetes picture, what nodes, pods, and clusters cost.
FAQs
Amazon EKS charges $0.10 per cluster per hour, about $73 a month, for a cluster on a Kubernetes version in standard support, before any workloads run. Once a version leaves standard support, 14 months after release, the cluster moves automatically to extended support at $0.60 per hour, about $438 a month. The larger cost is the compute underneath: EC2 nodes at their instance rates, Fargate pods per vCPU-hour and GB-hour, or EKS Auto Mode at the EC2 rate plus a management fee of roughly 12 percent.
EKS Auto Mode is the option in which AWS provisions, scales, patches, and upgrades the nodes for you, bundling Karpenter, the EBS CSI driver, the load balancer controller, and node lifecycle. It costs the standard cluster fee plus the EC2 instance price plus a management fee of about 12 percent on top of the on-demand instance rate, varying by instance type; for an m5.large in US East the fee is about $0.0115 an hour on a $0.096 instance.
Fargate bills each pod's allocated vCPU and memory per second, with a one-minute minimum, at $0.04048 per vCPU-hour and $0.004445 per GB-hour in US East on x86 (20 percent less on Arm), with no nodes to manage and 20 GB of ephemeral storage included per pod. There is no Fargate cluster fee beyond the $0.10-per-hour EKS control plane. It is simpler than nodes and costs more per unit at high, steady utilization.
When a cluster's Kubernetes version passes its 14-month standard support window, EKS continues to support it for a further 12 months at six times the control-plane fee, $0.60 per hour instead of $0.10, automatically. Version 1.33 entered extended support on July 29, 2026. Upgrading on a cadence that stays within standard support avoids the charge entirely.
Everything the workloads consume from other AWS services: EBS volumes behind persistent volume claims (gp3 about $0.08 per GB-month), load balancers (about $16 a month each plus usage and two or three public IPv4 addresses), NAT Gateways (about $33 a month plus $0.045 per GB), public IPv4 addresses ($0.005 per hour each), data transfer across zones and to the internet, CloudWatch logs and metrics, and ECR storage. These are usually larger than the control plane and often rival the nodes.
On the control plane, no: ECS is free, EKS is $73 a month per cluster. On compute they are the same, since both run on EC2 or Fargate at identical rates. The difference is usually small against the node bill, and the choice turns on whether you need Kubernetes itself, its ecosystem, portability, and tooling, or whether ECS's simpler model fits.