Google CloudGoogle Kubernetes EngineContainers

`google_container_cluster` cost estimation

A managed Kubernetes cluster on GKE. Two operating modes: Standard with per-cluster management fee, or Autopilot priced per pod.

A google_container_cluster is a managed Kubernetes cluster on GKE. GKE has two distinct operating modes with very different cost profiles.

Standard mode: you manage node pools (google_container_node_pool), much like EKS or AKS. The control plane charges $0.10/hour ($73/month) per cluster as a management fee. Worker nodes are regular Compute Engine VMs, billed at standard GCE rates with sustained-use and committed-use discounts applying.

Autopilot mode (enable_autopilot = true): GKE manages the nodes for you and bills per running pod by requested CPU, memory, and ephemeral storage. There's no separate control plane fee on Autopilot. Pod requests are billed per vCPU-hour, GB-hour, and GB-hour of ephemeral storage. For comparable workloads, Autopilot is usually 10-30% more expensive than well-utilized Standard, but you save engineering time on node management.

Standard mode is right when you need GPU pods, custom node OS, daemonsets, or have steady utilization. Autopilot is right when utilization is variable, you don't want to manage nodes, or you have few pods.

Additional costs apply for: load balancers (each Service of type LoadBalancer creates a google_compute_forwarding_rule), persistent volumes (google_compute_disk via the GKE CSI driver), and external IPs.

c3x reads the cluster mode and node pool configurations. Autopilot pods need pod request specs from the deployment manifests or usage file.

Terraform example

A minimal but realistic configuration that C3X can estimate.

resource "google_container_cluster" "production" {
  name     = "production"
  location = "us-central1"

  remove_default_node_pool = true
  initial_node_count       = 1

  release_channel {
    channel = "REGULAR"
  }
}

resource "google_container_node_pool" "general" {
  name       = "general-pool"
  cluster    = google_container_cluster.production.id
  node_count = 3

  node_config {
    machine_type = "e2-standard-4"
    disk_size_gb = 100
    disk_type    = "pd-balanced"
  }

  autoscaling {
    min_node_count = 2
    max_node_count = 10
  }
}

Pricing dimensions

What you actually pay for when you provision google_container_cluster.

Dimension	Unit	What's being charged
Standard cluster management fee	per cluster per hour	Flat hourly fee for the control plane on Standard mode. Free for the first zonal cluster per billing account. $0.10/hour ≈ $73/month
Node VMs (Standard mode)	per VM per hour	Worker nodes are GCE instances. Standard GCE pricing applies, including SUD/CUD.
Node disks	per GB-month	Boot and persistent disks for each node.
Autopilot pod CPU	per vCPU-hour	Billed by the requested CPU of each pod. Rounded to nearest 0.25 vCPU.
Autopilot pod memory	per GB-hour	Billed by the requested memory of each pod. Rounded to nearest 1 GB.
Autopilot ephemeral storage	per GB-hour	Billed by requested ephemeral-storage of each pod.

Optimization tips

Common ways to reduce google_container_cluster cost without changing the workload.

Take advantage of the free zonal cluster

$73/month

GKE includes one free zonal cluster per billing account. Use it for dev/staging to avoid the $73/month management fee on Standard mode.

Use Spot VMs for batch and stateless workloads

60-91% on compute

Spot VMs in GKE Standard cost 60-91% less than on-demand. Right for CI/CD runners, batch jobs, and stateless services that tolerate eviction.

Use Autopilot for low-utilization workloads

Workload-dependent

Standard mode pays for nodes even when pods underutilize them. Autopilot bills per pod request, so low-utilization clusters cost less. Crossover depends on average node utilization.

Buy Committed Use Discounts for Standard node pools

37-70%

1-year CUDs cut compute by 37%, 3-year by 70%. Right for production node pools with steady min_node_count.

FAQ

Standard or Autopilot for production?

Standard is more flexible (custom OS, GPUs, daemonsets, more control over node sizing). Autopilot trades flexibility for operational simplicity. Most production workloads still use Standard; Autopilot is a strong choice for small-team operations.

Does c3x include load balancer cost?

Yes. Each Kubernetes Service of type LoadBalancer creates a google_compute_forwarding_rule, which c3x estimates separately at $0.025/hour per rule.

Are GKE security updates free?

Yes. Node OS patching, control plane upgrades, and security fixes are included. The Premium tier of Anthos Service Mesh and other Anthos components have separate costs.

What about regional vs zonal clusters?

Regional clusters replicate the control plane across zones for HA. Both have the same $73/month management fee on Standard, but regional clusters typically have 3x the nodes (one per zone) by default. Zonal clusters can be cheaper for dev.