
Your final number shifts with the US region, control-plane tier, VM SKU and node count, operating system, storage type and provisioned size, and discounts such as Spot or Reserved pricing.
This guide covers the three tiers, node and storage pricing, recurring and hidden charges, and a practical way to build a realistic monthly budget. Prices below are USD retail estimates for East US as of November 18, 2026. Confirm them in the Azure Pricing Calculator before you commit.
Key Takeaways
- Control-plane cost runs from $0 (Free) to $73 (Standard) to $438 (Premium) per cluster per month.
- Worker nodes are Azure VMs and usually the biggest line item.
- Disks bill on provisioned size and performance tier, not on data stored.
- Pay for Standard or Premium only when SLA, support or version-lifecycle needs justify it.
How Much Does AKS Cost? (Pricing Overview)
AKS has no single monthly price. The bill is a control-plane fee, plus worker VM hours, plus OS and data disks, plus networking, monitoring, security, registry and backup. Microsoft's AKS FAQ confirms the control-plane fee is separate from node consumption.
Three misunderstandings cause most budget misses:
- Treating "free AKS" as a free cluster and leaving out node, disk, load balancer, monitoring and transfer charges.
- Estimating only application nodes while forgetting system node pools, minimum node counts, autoscaling headroom and always-on non-production environments.
- Using a generic online price without checking region, current SKU availability, billing model and usage assumptions.
Control-Plane Tiers
Fees are per cluster, not per node.
| Tier | Per cluster-hour | ~730-hour month | API-server availability | Support |
|---|---|---|---|---|
| Free | $0 | $0 | Internal goal above 99.5%, no financially backed SLA | Best-effort control plane |
| Standard | $0.10 | $73 | Financially backed 99.95% with availability zones, 99.9% without | Production-oriented |
| Premium | $0.60 | $438 | Same SLA as Standard | Long-term support plan |
Free suits development, testing and learning. Without a financially backed SLA, it's a poor fit for critical production systems.
Standard is the default production choice. The SLA covers the Kubernetes API server, not your application's uptime.
Premium adds long-term support: roughly 24 months of version support, measured from the version's AKS GA date rather than from when you enroll. It doesn't make nodes or storage cheaper. Choose it for lifecycle control, not performance.

Worker-Node Pricing
Each worker node is an Azure VM billed by SKU, region, operating system and time provisioned. GPU, memory-optimized and Windows nodes cost more. For a reference point, Linux Standard_D4s_v5 in East US lists at $0.192 per hour, per the Azure retail prices API.
Node count is driven by:
- Node-pool minimums and the system pool
- Autoscaler limits
- Pod CPU and memory requests
Requested capacity is not utilization. Oversized requests force the scheduler to reserve nodes that sit mostly idle, and you pay for the reservation.
Purchasing options compare like this:
| Model | Best for | Catch |
|---|---|---|
| Pay-as-you-go | Variable workloads | Highest unit rate |
| Reservations | Stable baseline | Commitment of one or three years |
| Savings Plans | Steady hourly spend | Hourly commitment is owed whether used or not |
| Spot | Interruptible work | No SLA; Azure can reclaim capacity |
Worker nodes set the floor for most AKS bills. Storage is the next major cost layer and the one teams underestimate most often.
Storage Pricing and Storage Tiers
AKS storage comes in four forms, each billed differently:
- OS disks: managed OS disks persist in Azure Storage, while ephemeral OS disks live on the host.
- Persistent volumes on Managed Disks: the Azure Disk CSI driver normally exposes ReadWriteOnce volumes to one node.
- Azure Files shares: shareable across nodes and pods.
- Ephemeral storage: tied to the node's lifecycle.
Disk types compare as follows, per Microsoft's disk documentation:
| Disk type | Max IOPS | Max throughput | Latency |
|---|---|---|---|
| Standard HDD | 2,000 | 500 MB/s | Writes mostly under 10 ms |
| Standard SSD | 6,000 | 750 MB/s | Single-digit ms |
| Premium SSD | 20,000 | 900 MB/s | Single-digit ms |
| Premium SSD v2 | 80,000 | 2,000 MB/s | Submillisecond |
| Ultra Disk | 400,000 | 10,000 MB/s | Consistently low submillisecond |
These are family maximums, not entry-tier performance. Premium SSD v2 includes 3,000 IOPS and 125 MB/s at no extra performance charge; anything above that adds cost.
Managed disks bill on provisioned size and performance tier, not on how full the filesystem is. Snapshots, unattached disks and retained volumes keep charging after the workload is gone.
Scenario-Based Cost Ranges
These East US scenarios use Standard_D4s_v5 at $0.192 per node-hour. They cover VM and control-plane cost only, not storage, networking, monitoring or licensing.
| Scenario | Nodes and runtime | Tier | Monthly subtotal |
|---|---|---|---|
| Scheduled development | 2-3 nodes, 160 hours each | Free | $61.44-$92.16 |
| Continuous production | 3-6 nodes, 730 hours each | Standard | $493.48-$913.96 |
| Larger LTS platform | 20-60 nodes, 730 hours each | Premium | $3,241.20-$8,847.60 |

Add persistent volumes, load balancers, monitoring and egress to each. All figures are illustrative; refresh them against current Azure rates.
Key Factors That Affect AKS Pricing and Total Cost Breakdown
Total cost is recurring infrastructure consumption plus any per-cluster management fee. The AKS line item alone tells you little.
Node Compute and Cluster Configuration
The recurring compute bill depends on VM family, vCPU-to-memory ratio, GPU needs, Linux versus Windows licensing, node count, system-pool overhead and hours provisioned. Windows nodes with eligible existing licenses can use Azure Hybrid Benefit to cut the VM cost.
Pod requests decide how many nodes you need. To shrink idle paid capacity without hurting SLOs:
- Rightsize requests and limits against observed usage.
- Let the Cluster Autoscaler add and remove nodes as pods become unschedulable.
- Add workload autoscaling for replicas.
Spot nodes fit batch jobs, CI/CD, development and fault-tolerant work. They need a secondary node pool, the spot taint with a matching toleration, and fallback capacity. The default eviction policy deletes evicted nodes. Keep critical services on regular or reserved capacity.
Persistent Storage and Storage Performance
Storage cost breaks down into provisioned capacity, performance tier, IOPS and throughput, redundancy, snapshots, backups and volume count. Full snapshots bill used data; incremental snapshots bill changes.
Match tier to measured need:
- Standard tiers for development or latency-tolerant workloads
- Premium SSD for transactional systems and databases
- Premium SSD v2 or Ultra only when you can show IOPS or latency requirements that justify them
Clean up as you go. Abandoned PVCs, unattached disks, old snapshots and oversized volumes all keep billing. A Retain reclaim policy keeps the underlying disk after the claim is deleted, and Microsoft's guidance on unattached disks notes they remain chargeable.
Networking, Monitoring, and Add-Ons
These meters sit outside the AKS line item:
- Standard Load Balancer: first five rules $0.025/hour, extra rules $0.01/rule-hour, $0.005/GB processed (Central US).
- Application Gateway: East US Standard_v2 is $0.246/hour fixed plus $0.008 per capacity-unit-hour.
- Public IPs and NAT Gateway: hourly resource charges, plus processed data for NAT.
- Egress: first 100 GB/month free, paid tiers after. Same-region zone-to-zone transfer is free; US inter-region transfer runs $0.02/GB.
- Container Insights and Log Analytics: ingestion and retention, with 31 days of retention typically included.
- Defender for Containers, Container Registry, Azure Backup for AKS: per vCore, per tier and storage, or per protected namespace.

One-time planning and migration effort is separate from this recurring consumption. Watch for periodic costs too: upgrades, backup snapshots, node replacements and commitment renewals.
Pricing Discounts and Usage Patterns
Match the purchase model to how you run capacity:
- Pay-as-you-go for variable demand
- Reservations and Savings Plans for a measured, stable baseline
- Spot for interruptible workloads
Discounts apply to underlying compute, not every AKS charge.
Base commitment size on actual utilization and your minimum sustained node count, never on peak or an over-provisioned baseline.
Schedule shutdowns for dev, test, staging and preview clusters. Stopping a cluster preserves state for up to 12 months, but storage keeps billing while nodes are off.
Low-Cost vs High-Cost AKS: What's the Difference and How Much Should You Budget?
Cheaper isn't automatically better. The right design balances availability, performance, resilience, operational effort and long-term cost.
Lower-Cost AKS Configuration
- Free tier for non-critical clusters
- Right-sized general-purpose nodes with autoscaling
- Standard-tier disks where latency allows
- Few add-ons, scheduled non-production hours
The trade-off is no financially backed API SLA and less headroom for spikes. Spot nodes cut cost further for eligible work, but pair them with taints, tolerations, fallback pools and retry or checkpoint logic.
Higher-Cost AKS Configuration
Higher bills typically come from:
- Standard or Premium control plane tiers
- Larger or GPU nodes and multiple clusters
- High-performance disks and multi-zone architecture
- Continuous observability, backups and heavy egress
You gain SLA-backed API availability, resilience and support. Justify Premium on version-support needs rather than assuming it improves application performance.
How to Estimate the Right Budget
- Gather workload requirements: applications, pod requests, baseline and peak replicas, availability targets, data growth, latency, compliance and interruption tolerance.
- Pick a US region and document assumptions: SKU, node count, system pool, OS, runtime hours, tier, storage type and size, redundancy, traffic, log volume, add-ons.
- Build the estimate in the Azure Pricing Calculator: use Retail Prices API unit rates, then verify each line against the official pricing page.
- Model three scenarios: steady state, expected peak, and optimized or discounted. Mark what scales with usage and what is fixed per cluster or resource.
- Add contingency, then compare the forecast against Azure Cost Management after deployment.

With Standard or Premium on an Enterprise Agreement or Microsoft Customer Agreement, AKS cost analysis breaks out idle, system and unallocated charges by namespace.
What Most People Miss When It Comes to AKS Cost
The calculator is only as accurate as your inputs. It won't reveal idle capacity from oversized pod requests, system-pool overhead, minimum node counts, or slow scale-down.
Commonly overlooked charges:
- Provisioned-but-unused disk capacity and unattached disks
- Retained snapshots and unneeded performance tiers
- Log ingestion and retention
- Public IPs and load-balancer rules
- Internet egress and extra clusters
Storage deserves a regular review, not an annual one. Lucidity's internal figures, drawn from 600+ Assessments covering 100+ PB, put average enterprise disk utilization at 30%. That's Lucidity's cross-cloud data, not an AKS-specific number.
Lucidity is an optional way to get visibility here:
- Free Assessment scans block storage in about five minutes with no installation
- Lumen scores each disk's tier on IOPS, throughput, latency and cost, and can recommend moves such as Premium SSD to Standard SSD
- AutoScaler expands and shrinks eligible disks with zero downtime
Keep Azure Cost Management as your source for billing validation.
Conclusion
AKS pricing has three layers: the control-plane tier, Azure VM worker nodes, and supporting resources such as storage, networking, monitoring, and backups.
The most reliable budget uses current US-region prices and spells out node, storage, runtime, scaling, and add-on assumptions instead of quoting one universal figure.
The right AKS spend comes from a configuration that meets your reliability and performance needs while cutting idle nodes, over-provisioned disks, unnecessary add-ons, and poorly sized commitments.
Frequently Asked Questions
How much does Azure Kubernetes Service (AKS) cost?
Total cost combines the control-plane tier ($0, $73, or $438 per month in East US) with worker VMs, storage, networking, monitoring, and other services. Use current regional pricing rather than a single universal total.
How do I check Azure AKS pricing?
Use the Azure Pricing Calculator with the official AKS, VM, Managed Disks, bandwidth, and monitoring pricing pages. Enter region, VM SKU, OS, node count, hours, tier, disk type and size, traffic, and log volume.
What is included in the AKS free tier?
The free tier removes the control-plane management fee. Worker nodes, storage, networking, monitoring, and data transfer still bill, and there is no financially backed uptime SLA.
What do AKS nodes cost?
Pricing depends on VM SKU, region, operating system, node count, runtime and billing model. Include system nodes and autoscaler maximums in the estimate.
How much does storage cost in AKS?
It depends on the backing service, provisioned capacity, performance tier, redundancy and snapshots. Managed Disks, Azure Files and ephemeral storage each bill differently.
What costs should I include beyond AKS node pricing?
Add control-plane fees, persistent volumes, OS disks, load balancers, public IPs, egress, and inter-region traffic. Also factor in Azure Monitor and Log Analytics, security add-ons, container registries, backups, and retained resources.