Kubernetes and Container Cost Tools
Free tools that price Kubernetes clusters, GKE Autopilot, Fargate and Cloud Run from what your workloads request.
4 free tools · no login
Tools
Kubernetes Cost Calculator
Estimate your cluster's cost on EKS, GKE or AKS, including load balancers, disks and data transfer.
Fargate vs EC2 Calculator
Compare running containers on Fargate against the EC2 instances you would need to buy for the same work.
Cloud Run Cost Calculator
Compare the two Cloud Run billing models on your own traffic and see which one costs less each month.
GKE Cost Check & Autopilot Advisor
Check your cluster's capacity costs and see whether Autopilot is worth investigating.
What drives the cost
A Kubernetes estimate that starts from node count is wrong in two ways.
Too low: a cluster is more than its nodes
Besides nodes, a cluster pays a control-plane fee, a load balancer per LoadBalancer Service, a disk per node, and egress and cross-zone traffic. On a small cluster these can cost more than the compute.
Too high: Pods request more than they use
Nodes are sized to Pod requests, so you pay for requests, not usage. Most clusters use a third to half of what they request: the largest source of Kubernetes waste.
How the calculator works
The calculator starts from Pod requests, subtracts system reservations and daemonsets, packs the Pods onto nodes, and only then prices the result.
EKS, GKE and AKS: similar fees
The standard control-plane fee is $0.10 per cluster-hour on each. The AKS Free tier costs nothing; EKS or GKE extended support, and AKS Premium, cost $0.60. Node prices matter more than the fee.
Fargate and Cloud Run: no nodes, higher rates
Fargate and Cloud Run remove the node layer and its unrequested capacity. You still pay for the size you configure, at a higher unit price. At 30% utilization that premium often costs less than the waste it removes; at 70% it does not.
Other categories
Requests drift. So does the gap between requested and used.
Finitizer tracks requested versus actual capacity per namespace and workload over time, attributes cluster cost back to the teams causing it, and flags version cliffs and idle node groups before they reach an invoice.
See how Finitizer works