The Ultimate Guide to EKS Pricing: Nodes, Control Plane, Storage, Networking

EKS pricing is more than the $0.10/hour control-plane fee advertised on the homepage. Use this guide as a checklist when planning new clusters or auditing existing ones.

1. Control plane and add-ons

• EKS control plane: $0.10/hour (~$73/month) per cluster.
• EKS Add-ons: Managed versions of CoreDNS, kube-proxy, VPC CNI. The add-ons themselves are free, but the resources they consume (CPU, memory, logs) cost money.
• Cluster logging: Shipping API server, audit, or scheduler logs to CloudWatch incurs ingestion and storage fees.

2. Worker capacity

Option	How you pay	Notes
Managed node groups	EC2 pricing + EBS	AWS handles upgrades/rollouts
Self-managed nodes	EC2 pricing + EBS	More control, more ops
Fargate profiles	Per vCPU-second + GiB-second	Great for spiky or small workloads
Bottlerocket	Same EC2 pricing	Lower footprint, faster patching

Tip: Mix purchase options—reserved instances for steady workloads, spot for bursty ones, on-demand for the rest. ClusterCost helps simulate savings-plan coverage.

3. Storage

• EBS volumes: gp3/gp2/io1. Remember to factor in provisioned throughput on gp3.
• Snapshots: Charged per GB-month until deleted.
• EFS/FSx: Pay-per-GB stored plus throughput tiers.
• Ephemeral storage (ephemeral-storage resource): Included on nodes but can force upgrades if pods exceed limits.

4. Networking

• Load balancers: ALB/NLB hourly + per-LCU charges.
• NAT gateways: $0.048/hour + per-GB processed. Shared NAT gateways can become six-figure line items.
• Data transfer: Intra-AZ free, cross-AZ and internet egress billed per GB.
• PrivateLink/Transit Gateway: Additional hourly and data charges when connecting environments.

5. Observability + platform tooling

• CloudWatch metrics/logs (ingest and storage).
• X-Ray/OTel collectors running in the cluster.
• Container registry egress (ECR data transfer).

While not “EKS fees” per se, they scale with cluster size and should be budgeted together.

Forecasting template

1. Estimate node hours per instance family.
2. Add 20% buffer for autoscaler headroom.
3. Include storage multipliers (1.3× for stateful clusters).
4. Model networking based on expected request/traffic volume.
5. Layer on tooling (monitoring, logging) as a % of node spend (commonly 8–12%).

ClusterCost’s planner automates this math: plug in desired capacity, mix of purchase options, and traffic assumptions to see monthly cost before rolling out.

Optimization checklist

• Turn on control-plane logging only for environments that need it.
• Standardize on gp3 volumes with tuned throughput.
• Allocate NAT and load balancer costs to the teams who need them to drive responsible usage.
• Use Karpenter or node rightsizing to match instance types with workloads automatically.

Armed with this breakdown, you can walk into any budget conversation with confidence and demonstrate exactly how EKS spend maps to business value.***