The question I get asked most often: "Should we buy Reserved Instances or Savings Plans?"
The honest answer: it depends. But after this guide, you will know exactly which to choose for your workload.
Savings Plans if you want flexibility and are primarily running EC2 or Fargate.
Reserved Instances if you have predictable, static workloads and want maximum savings on specific instance families.
But the real answer is more nuanced. Let us dig in.
RIs are the original AWS commitment model. You commit to a specific:
In exchange, you get up to 72% off On-Demand pricing.
Types of RIs:
Launched in 2019, Savings Plans offer more flexibility. You commit to a dollar amount per hour (e.g., $10/hour) rather than specific instances.
Types of Savings Plans:
| Factor | Reserved Instances | Savings Plans |
|--------|-------------------|---------------|
| Max Discount | Up to 72% | Up to 66% |
| Flexibility | Low (Standard) / Medium (Convertible) | High |
| Applies To | EC2, RDS, ElastiCache, Redshift, OpenSearch | EC2, Fargate, Lambda |
| Commitment | Instance-level | Dollar amount |
| Management | More complex | Simpler |
1. Database workloads
RDS, ElastiCache, and Redshift do not support Savings Plans. If you have production databases running 24/7, RIs are your only option for commitments.
2. Stable, predictable EC2 usage
If you know you will run 10 x m5.xlarge instances in us-east-1 for the next 3 years, Standard RIs give you the maximum discount.
3. You want the absolute lowest price
Standard RIs offer up to 72% off, compared to 66% for Savings Plans. That 6% difference matters at scale.
1. You are modernizing your architecture
Moving to containers? Considering Fargate? Migrating to Graviton? Compute Savings Plans flex with you.
2. Your instance mix changes frequently
If you are constantly right-sizing, experimenting with instance types, or running varied workloads, Savings Plans adapt.
3. You want simplicity
Savings Plans are easier to manage. One commitment covers multiple services and instance types.
4. You use Fargate or Lambda significantly
These services ONLY benefit from Savings Plans, not RIs.
Most mature FinOps practices use both:
This gives you maximum savings on predictable database workloads while maintaining flexibility for compute.
Step 1: Look at your last 3 months of On-Demand usage
Step 2: Identify the stable baseline (minimum consistent usage)
Step 3: Commit to 70-80% of that baseline
Why not 100%? Because usage fluctuates. A 70-80% commitment gives you high coverage without over-committing.
3-year terms offer better discounts but carry risk:
My recommendation: Start with 1-year terms. Only go 3-year when you have high confidence in long-term usage patterns.
Buying commitments that exceed your usage means you are paying for capacity you do not use. Always leave headroom.
Being too conservative leaves savings on the table. If you have been running stable workloads for 6+ months, you have data to commit confidently.
Track your commitment utilization weekly. If it drops below 80%, you over-committed. Adjust next time.
Need help optimizing your commitment strategy? That is exactly what we do at FinOps Fanatics. Book a discovery call and we will show you what you are leaving on the table.
