Kubernetes Pricing Explained (What You Pay)
The control plane fee everyone quotes is the smallest line on your bill. Here is where Kubernetes money actually goes, with real numbers and the traps that quietly inflate the total.
Key takeaways
- The control plane fee (~$73/month on EKS and GKE, free on AKS) is under 5% of a typical Kubernetes bill; worker nodes are 60-70% and deserve your attention instead.
- A small 3-node production cluster runs roughly $250-300/month all-in; cost scales nearly linearly with node count.
- The real surprises are storage billed on provisioned (not used) capacity, ~$22/month per load balancer, and cross-zone or internet egress that can add $50-150+/month.
- Spot or preemptible nodes cut compute 60-90% for fault-tolerant work; right-sizing over-requested pods is the other big lever.
- Self-managed Kubernetes can lower the invoice but typically needs ~3x the platform engineers, so managed is usually cheaper once you count people.
The Five Things You Actually Pay For
Kubernetes pricing confuses people because there is no single "Kubernetes price." Kubernetes itself is open-source and free. What you pay for is the infrastructure it runs on and, usually, a managed control plane to keep it healthy. Your bill is the sum of five buckets, not one number.
Roughly speaking, here is how a typical managed cluster bill splits: worker node compute is 60-70%, storage and networking together are another 20-30%, and the managed control plane fee is under 5%. That last figure surprises people, because the control plane fee is the number every comparison article leads with. It is the least important line on your invoice.
- Control plane: the managed Kubernetes "brain" (API server, scheduler, etcd) the provider runs for you
- Worker nodes: the VMs or bare-metal servers your containers actually run on (the big cost)
- Storage: persistent volumes for databases, uploads, and stateful apps
- Networking: load balancers, NAT gateways, and data transfer (the sneaky cost)
- Add-ons: monitoring, logging, container registry, and ingress controllers
Control Plane Fees: The Number Everyone Quotes
The control plane is the managed part of managed Kubernetes. The provider runs and patches the API server, scheduler, and etcd datastore so you don't have to. Pricing here is refreshingly simple and nearly identical across the big three clouds.
Amazon EKS and Google GKE Standard both charge about $0.10 per cluster per hour, which works out to roughly $73 per month before a single container runs. Azure AKS charges nothing for the control plane on its free tier and $0.10/hour for the SLA-backed Standard tier. GKE Autopilot waives the control plane fee but bills you per-pod for the resources you request instead.
One real trap worth knowing: EKS charges $0.60/hour, about $438/month, for clusters running a Kubernetes version that has aged into "extended support." Letting your version drift is an easy way to 6x your control plane cost. Keep clusters on a supported version and this line stays at $73.
Worker Nodes: Where the Money Actually Goes
Worker nodes are the VMs or bare-metal machines that run your pods, and they are 60-70% of a typical bill. This is where you should spend your optimization energy, not on the $73 control plane fee.
A concrete example: a small production cluster of three mid-size nodes (think 2 vCPU / 8 GB each) runs roughly $250-300/month all-in on standard on-demand pricing. Three such nodes alone are around $210/month; the rest is storage, a load balancer, and transfer. Node cost scales linearly, so the same cluster at 30 nodes is roughly 10x the compute.
Two levers move this number the most. First, right-sizing: most clusters request far more CPU and memory than pods actually use, and you pay for the request, not the usage. Second, spot or preemptible instances for fault-tolerant workloads, which run 60-90% cheaper than on-demand. Bare-metal nodes are a third lever; they remove the virtualization tax and per-VM markup entirely, which is why high-density workloads often land on dedicated hardware.
Storage, Load Balancers, and the Egress Trap
The remaining 20-30% of your bill hides in storage and networking, and this is where surprise charges live.
Persistent volumes bill for what you provision, not what you use. A 100 GB volume backing a database that only holds 15 GB still costs you the full 100 GB every month. Block storage runs around $0.08 per GB-month, so a 100 GB volume is about $8/month, and they add up fast across stateful services.
Load balancers are about $22/month each, plus data processing, and clusters rarely stop at one; every externally exposed service tends to spawn another. The genuinely sneaky line is cross-zone data transfer. Kubernetes spreads pods across availability zones for resilience, and many clouds charge about $0.01 per GB when those pods talk to each other across zones. For a chatty microservices app, cross-AZ traffic alone can quietly add $50-150/month. Egress to the public internet is priced higher still and is the single most common bill shock.
Managed vs. Self-Managed: The Total Cost Picture
The cheapest-looking invoice is not always the cheapest cluster. Running Kubernetes yourself on bare metal removes the per-cluster control plane fee and the per-VM markup, which can dramatically cut the raw infrastructure number. But you now own upgrades, security patching, etcd backups, and 3 a.m. incidents.
Industry surveys consistently show self-hosted Kubernetes typically needs around three platform engineers to run safely, versus roughly one for a managed offering. At loaded salaries, that staffing delta dwarfs a $73/month control plane fee. The honest rule: self-managed wins on total cost only when you already have deep SRE bench strength and enough scale to amortize it. For most teams, managed Kubernetes is cheaper once you count the people.
The middle path many teams land on is managed control plane plus cost-optimized nodes, mixing on-demand for baseline, spot for burst, and bare-metal for predictable high-density workloads. That captures most of the savings without taking on the operational burden of running etcd yourself.
How to Estimate Your Own Bill
You can sketch a realistic monthly estimate in five lines without a calculator-heavy spreadsheet. Start from the workload, not the price list.
Add a healthy buffer. Real bills run 20-40% over a naive node-only estimate once idle headroom, monitoring add-ons, and inter-zone traffic are counted. Estimating only nodes is the most common way teams underbudget.
If you'd rather not assemble and price every line yourself, transparent managed plans bundle the control plane, predictable node pricing, and free migration so the total is knowable up front. You can see Kubernetes plans with all-in pricing on our managed Kubernetes page, including bare-metal node options for teams that want the density without running the control plane themselves.
- Control plane: ~$73/month per cluster (or free on AKS / self-managed)
- Nodes: (number of nodes) x (instance price); this is 60-70% of the total
- Storage: total provisioned GB x ~$0.08, not the GB you actually use
- Networking: ~$22/month per load balancer + cross-zone and egress transfer
- Add-ons + buffer: add 20-40% for monitoring, logging, and idle headroom
FAQ
How much does Kubernetes cost per month?
Kubernetes itself is free and open-source; you pay for infrastructure. A small managed production cluster of three mid-size nodes runs roughly $250-300/month all-in. The managed control plane is about $73/month on EKS and GKE (free on AKS), but worker nodes, storage, load balancers, and data transfer make up the bulk of the bill.
Why is the control plane fee so small compared to my bill?
The control plane fee (~$73/month) only covers the managed API server, scheduler, and etcd. Everything that does actual work, your worker nodes (60-70% of the bill), plus storage, load balancers, and network egress, is billed separately. Optimizing nodes saves far more than worrying about the flat control plane fee.
What are the hidden costs in Kubernetes pricing?
The three biggest surprises are persistent volumes billed on provisioned rather than used capacity, load balancers at around $22/month each that multiply as you expose more services, and cross-zone or public-internet data transfer that can add $50-150+/month for chatty applications. Idle, over-provisioned nodes are another quiet drain.
Is self-managed Kubernetes cheaper than managed?
On the invoice, sometimes yes, since you skip the control plane fee and per-VM markup. But self-managed clusters typically need around three platform engineers versus roughly one for managed, and that salary cost usually outweighs the savings. Self-managed wins on total cost mainly at large scale with an experienced SRE team.
Managed Kubernetes vs Self-Managed: Which to Pick
Both run the same upstream Kubernetes — the difference is who operates the control plane, handles upgrades, and gets paged at 2 a.m. Here's an honest look at the cost, control, and operational trade-offs of managed vs. self-managed Kubernetes, so you can match the model to your team instead of your ego.
GuideTypes of Web Hosting: Shared vs VPS vs Cloud vs Dedicated
Four hosting models, four very different trade-offs in performance, control, and cost. Here's how shared, VPS, cloud, and dedicated actually differ — with real numbers — so you can match the plan to your workload instead of overpaying or outgrowing it in six months.
GuideHow Much Does Web Hosting Cost in 2026?
A clear, no-hype breakdown of what web hosting actually costs in 2026 by hosting type, why the sticker price rarely matches the renewal bill, and how to budget for the plan you really need.