To use commitments with Kubernetes well, you commit to the part of your node fleet that is genuinely steady and leave the variable part on flexible or interruptible capacity. Because clusters autoscale, the node count moves throughout the day, so the right commitment covers the floor the fleet rarely drops below, not the peak. Sizing to the floor captures the discount on the capacity you always run while avoiding paying for reserved capacity during the troughs when the autoscaler has scaled the fleet down.
This article is part of our Kubernetes and container cost cluster. For the full picture, start with our complete guide to Kubernetes cost optimization, the pillar this piece links up to. Commit only after rightsizing, because a baseline built on bloated requests over-commits, as explained in the cost of over-provisioned Kubernetes clusters.
Step 1 · Rightsize before you commit
Commitments lock in whatever fleet you have for one to three years, so committing on an over-provisioned cluster locks in the waste. Always rightsize requests and consolidate first so the baseline reflects the real, efficient node count, then size commitments to that. The sequence is non-negotiable: See and Cut before you Lock. The rightsizing work is in how to rightsize Kubernetes requests and limits, and the idle removal in how to reduce idle capacity in Kubernetes.
Step 2 · Find the steady floor
Plot your node-hours by instance family over several weeks and identify the floor, the level of capacity the fleet essentially never drops below. That floor is the safe commitment target because you will consume it regardless of how the autoscaler flexes. Commit to roughly that floor, leaving headroom so a future efficiency gain does not strand the commitment. The variable layer above the floor stays on on-demand and Spot, where it can scale freely without wasting reserved capacity.
Unsure how much of your cluster to commit?
Our cost audit analyzes your node-hour baseline, identifies the steady floor by instance family, and recommends a commitment plan sized to consume, not to waste. On the performance model, you pay only from realized savings. No savings, no fee.
Book a cloud cost audit →Step 3 · Choose flexible commitment instruments
Each cloud offers commitments that differ in flexibility. Spend-based or flexible instruments, such as compute savings plans and spend-based committed use discounts, apply across instance families and adapt as the autoscaler changes node shapes, which suits dynamic clusters far better than rigid per-instance reservations. Resource-based commitments are deeper but lock you to a family, so they fit only the most stable part of the floor. Favor flexible instruments for the cluster, because the autoscaler will change instance choices over the commitment term.
Step 4 · Layer Spot above the committed floor
Above the committed steady floor, the variable and interruptible capacity belongs on Spot, which delivers the deepest discount of all for workloads that tolerate disruption. This three-layer model, committed floor plus on-demand buffer plus Spot for the flexible top, captures discount at every level without over-committing. The Spot mechanics are in how to use Spot instances for Kubernetes workloads, and the autoscaler that mixes these capacity types is covered in Cluster Autoscaler vs Karpenter for cost.
Step 5 · Monitor coverage and utilization
After committing, watch two numbers: commitment utilization, the share of your reserved capacity actually consumed, and coverage, the share of eligible spend that a commitment discounts. Aim for high utilization so you are not paying for unused reservations, and grow coverage only as the steady floor proves stable. Review quarterly, because a cluster that gets more efficient can outgrow its commitments and a growing one can justify more.
| Capacity layer | How variable | Best instrument |
|---|---|---|
| Steady floor | Always running | Flexible commitment |
| Daily variation | Predictable peaks | On-demand buffer |
| Burst and batch | Interruptible | Spot |
| Stable single family | Rarely changes shape | Resource-based commitment |
Commitment instruments and their flexibility above reflect each provider as of May 2026 and change. Verify the current commitment options, terms, and how they apply to managed Kubernetes node capacity in each provider's documentation before purchasing.
The Kubernetes Cost Optimization Handbook includes the commitment-sizing worksheet and the three-layer capacity model behind this article. It is the downloadable companion.
The short version
Use commitments with Kubernetes by rightsizing first, finding the steady floor your fleet never drops below, committing to that floor with flexible instruments, and leaving the variable top on on-demand and Spot. Monitor utilization and coverage and review quarterly. When you want the floor measured and a commitment plan sized to consume rather than waste, that is what our rightsizing and waste elimination service delivers.