Explanation
Why the waste happens and who it affects.
When that usage is not covered by reserved DB instances or Database Savings Plans, every instance-hour is billed at the On-Demand rate. The gap is often large because database commitments are managed separately from EC2 Savings Plans, which do not apply to RDS or Aurora.
Coverage is frequently missing for practical reasons: teams expect to change engine, instance family or Region, reserved DB instances are tied to engine, instance class type and Region, reservations for SQL Server and Oracle License Included are not size flexible, and expiring reservations quietly return usage to On-Demand. Database Savings Plans, introduced in December 2025, remove some of that lock-in by applying across engines, instance families, sizes, Regions and serverless usage, and commitment plans written before then do not account for them.
Billing model
The pricing dimensions that drive this cost.
RDS and Aurora instance usage is billed at On-Demand rates unless a commitment applies, and commitments discount instance-hours only.
- On-Demand DB instance-hours
- The full published rate per DB instance-hour for the instance class, engine and deployment option
- Reserved DB instances
- 1- or 3-year commitments tied to Region, engine and instance class type, billed for every hour of the term regardless of usage; No Upfront is available only as a 1-year reservation
- Size flexibility
- Reserved DB instances apply across sizes within the same instance class type for most engines, but not for RDS for SQL Server or RDS for Oracle License Included
- Database Savings Plans
- A 1-year, no upfront dollar-per-hour commitment that applies to Generation 7 and newer instances and to serverless usage across engines, families, sizes and Regions
- What is not discounted
- Storage, backup storage and I/O charges are not discounted by reserved DB instances
How to detect
5 checks to find it in your estate.
- Review the Trusted Advisor check Amazon RDS Reserved Instance purchase recommendations, sourced from Cost Optimization Hub, and the Cost Explorer reservation recommendations for Amazon RDS using a 30- or 60-day lookback
- Group RDS spend by Purchase Option in Cost Explorer, and review reservation and Savings Plans coverage, to find instance-hours consistently billed On-Demand
- Identify DB instances and Aurora clusters that have run continuously on the same instance class for the lookback period with no matching reservation, including reader instances, which need their own coverage
- Review Database Savings Plans purchase recommendations for Aurora Serverless ACU usage and for fleets that are expected to change instance family, engine or Region
- List existing reserved DB instances with their end dates, and check Aurora clusters that moved to I/O-Optimized, which consumes 30% more normalized units per hour than Aurora Standard and may now be only partly covered
How to fix
6 ways to remove the waste.
- Cover stable instances that will stay on the same engine and instance class type for the term with size-flexible reserved DB instances, choosing the term and payment option that matches the confidence in the baseline
- Use Database Savings Plans for baselines that may change engine, instance family, size or Region, move to Aurora Serverless, or run on SQL Server or Oracle License Included where reservations are not size flexible; confirm the instances are Generation 7 or newer
- Assign each workload to one commitment type, since Database Savings Plans cannot be combined with RDS reserved instances on the same workload
- Right-size and complete planned Graviton or generation upgrades before purchasing, because a reserved DB instance for one instance class type does not apply to another (for example db.r6i to db.r7g)
- For Aurora I/O-Optimized clusters, buy about 30% additional reserved capacity relative to the equivalent Aurora Standard usage to keep full coverage
- Review coverage, utilization and upcoming expirations on a regular schedule and size new commitments to the minimum steady baseline
Documentation
Vendor references for pricing and configuration.