Savings Plan
A flexible AWS pricing model where you commit to a minimum spend per hour for 1 or 3 years in exchange for up to 72% discount on compute services including EC2, Lambda, and Fargate.
A Savings Plan is a flexible pricing commitment that gives you up to 72% discount on AWS compute in exchange for committing to a minimum spend per hour for 1 or 3 years. It is the modern replacement for Reserved Instances with more flexibility.
How It Works
You commit: "I will spend at least $10.00 per hour on AWS compute for 1 year"AWS gives: up to 72% discount on your first $10 of compute per hourUsage above $10/hour: charged at standard On-Demand rates No specific instance type to choose. No specific region to lock in (for Compute SP).Just a dollar commitment — any matching compute usage gets the discount automatically.Two Types of Savings Plans
Compute Savings Plans (recommended for most teams): Applies to: EC2 (any instance family, any size, any region, any OS), Lambda, Fargate Maximum flexibility: change instance type, change region → discount follows automatically Discount: up to 66% Example: commit $5/hr Compute SP today on m5.large in ap-south-1 Next month switch to c6i.xlarge in us-east-1 → same discount applies automatically EC2 Instance Savings Plans (deeper discount, less flexibility): Applies to: one specific instance family in one specific region Example: commit to m5 family in ap-south-1 only Cannot change family or region without losing the discount Discount: up to 72%Payment Options
No Upfront: pay monthly, lower discountPartial Upfront: pay part now, monthly remainder, medium discountAll Upfront: pay everything now, maximum discount 3-year commitment + All Upfront = maximum possible savings (up to 72%)1-year commitment + No Upfront = minimum commitment, still significant savings (~40%)Savings Plans vs Reserved Instances
| Savings Plans | Reserved Instances | |
|---|---|---|
| Commitment | Dollar amount per hour | Specific instance type |
| Flexibility | High (Compute) / Medium (EC2) | Low — locked to type and region |
| Applies to | EC2 + Lambda + Fargate | EC2 only (Standard RI) |
| Management | Automatic — applies to matching usage | Manual — assign to specific instances |
| Convertible RI | N/A | Change instance type (lower discount) |
How to Size Your Commitment
Cost Explorer → Savings Plans → RecommendationsSelect: 1-year or 3-year, No Upfront or All UpfrontAWS analyses your last 7-30 days of usageRecommends: exact commitment amount and projected annual savings Start conservative — commit to your steady baseline usage, not peak usage.Spiky peak usage on top of the commitment: charged at On-Demand rates.Cost Anomaly Detection
Once you have Savings Plans, monitor utilisation. Low utilisation means you are paying for commitment you are not using:
Cost Management → Savings Plans → Utilisation reportTarget: 100% utilisation of committed amountBelow 100%: you committed more than you are using → adjust on renewalTipStart with Compute Savings Plans rather than EC2 Instance Savings Plans. The flexibility to change instance types and regions without losing the discount is almost always worth the slightly lower discount rate. Only switch to EC2 Instance Savings Plans if you are absolutely certain about your long-term instance family and region choices.
RememberSavings Plans cover only compute charges (EC2 instance hours, Lambda duration, Fargate task hours). They do not cover data transfer costs, EBS storage, RDS, or any other AWS service. Those are priced separately.
Frequently Asked Questions
How is a Savings Plan different from a traditional Reserved Instance?
A Reserved Instance commits to a specific instance type, region, and often a specific family, whereas a Savings Plan commits to a dollar amount of compute spend per hour and automatically applies the discount across whichever eligible usage you actually run — different instance families, sizes, regions (for Compute Savings Plans), or even across EC2, Fargate, and Lambda. This flexibility comes at a slightly lower maximum discount than an equivalent-term Reserved Instance locked to one exact configuration.
What's the financial risk of over-committing to a Savings Plan?
The commitment is a fixed hourly spend for the full 1- or 3-year term regardless of whether you actually use that much compute — if your workload shrinks (a service gets decommissioned, you migrate to serverless, usage drops), you're still billed for the committed amount, turning a savings mechanism into a fixed cost you can't reduce. The standard practice is committing conservatively based on your stable baseline usage, not peak usage, and layering on-demand or spot for the variable portion above that baseline.