Updated 25 October 2026 · Published by Zoorik
An AWS Savings Plan is a commitment to spend a set dollar amount per hour on compute, for one or three years. Usage it covers costs less than the on-demand rate. The Azure savings plan for compute works the same way.
An example from AWS
AWS’s example uses made-up rates. Four r5.4xlarge instances run for an hour, each $1.00 on-demand or $0.70 on a Compute Savings Plan.
A $2.00-an-hour plan covers about 2.9 of them. The other 1.1 are billed on-demand, for $1.14.
Savings plan vs reserved instance
- An AWS Compute Savings Plan takes up to 66% off on-demand, for any instance family, size or region.
- An EC2 Instance Savings Plan takes up to 72% off, for one family in one region.
These are AWS’s October 2026 figures, and prices change. Reserved Instances reach the same top discounts, but tie you to instance setups, not a dollar amount. Azure Reservations also need a set family, size and region.
Why it matters for your bill
- You pay the full commitment every hour, used or not, and can’t change it later.
- Spare nodes raise what you’d commit to, so reduce cloud costs first.
- On EKS, a plan covers your EC2 nodes, but not EKS’s own charges.
- On AKS, it covers the virtual machines behind your nodes, if Microsoft lists their series as eligible.
- Coral rebalances CPU and memory requests on AKS and EKS to match real use, and drains spare nodes so your autoscaler can remove them.
- The free assessment works out your own figure.
Read next: How to reduce cloud costs: what moves the bill
Sources: AWS, “Savings Plans types” (opens in a new tab); AWS, “Compute Savings Plans and Reserved Instances” (opens in a new tab); AWS, “Understanding how Savings Plans apply to your usage” (opens in a new tab); Microsoft Azure, “Azure Savings Plans” (opens in a new tab).