Committing to AWS usage via Reserved Instances (RIs) or Savings Plans is the single highest-impact lever most AWS customers can pull to reduce costs. The potential savings are substantial — up to 72% off on-demand pricing — but the commitment comes with real financial risk if you make the wrong choices. Locked into the wrong instance type as your architecture evolves, or committed to more than you use, and discounts become expensive mistakes.

At $15,000/month in EC2 spend, a 1-year Compute Savings Plan covering $12,000/month of that spend at a 33% discount saves $47,520/year. That's real, immediate, low-risk savings — but only if you set it up correctly. This guide gives you the full picture to make confident purchasing decisions.

72%
Max discount on 3-yr All Upfront Standard RI
66%
Max discount on 3-yr Compute Savings Plan
40%
Typical 1-yr Compute Savings Plan discount

Understanding the Commitment Products

AWS offers several distinct commitment products. Understanding what each covers and how they differ is essential to making smart purchasing decisions.

EC2 Reserved Instances

The original AWS commitment product. You commit to a specific EC2 instance family and region (Convertible RIs) or a specific instance type, region, OS, and tenancy (Standard RIs). In exchange, AWS discounts that instance's hourly rate.

Compute Savings Plans

The most flexible commitment product. You commit to a dollar amount of compute usage per hour (e.g., $10/hr). AWS automatically applies the discount to any qualifying usage, in any region, on any instance family, for EC2, Fargate, and Lambda.

EC2 Instance Savings Plans

A hybrid between Standard RIs and Compute Savings Plans. You commit to a specific EC2 instance family in a specific region (e.g., m5 in us-east-1), but the commitment is flexible within that family across any size and OS.

RDS Reserved Instances

Reserved Instances for Amazon RDS. Committed to a specific DB engine (PostgreSQL, MySQL, Aurora, etc.), instance class, and region. Can save up to 69% vs on-demand RDS pricing.

ElastiCache and Other Service RIs

Reserved Instances are also available for ElastiCache, Redshift, OpenSearch, and other services. The mechanics are the same — commit for 1 or 3 years, get a discount. Each service has its own RI product with service-specific flexibility rules.

Head-to-Head Comparison

Product Services Covered Max Discount (1yr) Max Discount (3yr) Flexibility Best Use Case
Compute Savings Plan EC2, Fargate, Lambda 40% 66% Any instance type/family/region General EC2 + mixed workloads
EC2 Instance Savings Plan EC2 only Up to 72% Up to 72% Any size within one family + region Stable instance family, still resizing
Standard RI EC2 only 40% 72% None — exact type/OS/region locked Frozen instance configurations
Convertible RI EC2 only 31% 54% Exchange to other types in same family Migrating between generations (m5 → m7g)
RDS RI RDS only 40% 69% Engine + class locked Stable production databases

Real-World Savings Example: $15,000/Month EC2 Spend

Let's work through a concrete example. Your team spends $15,000/month on EC2 on-demand, distributed across m5 and c5 instances in us-east-1. You've just completed a rightsizing exercise and your fleet is now correctly sized. How should you structure your commitments?

Approach 1: 1-Year Compute Savings Plan, No Upfront
Commit to $10/hr ($7,200/month at the discounted rate, covering ~$10,500/month of on-demand spend).
Discount: 33%
Annual savings: ~$3,960/month × 12 = $47,520/year
Risk: Low. If your architecture changes, the Savings Plan covers any EC2 usage automatically.

Approach 2: 1-Year EC2 Instance Savings Plan (m5 family, us-east-1) + 1-Year Compute Savings Plan for the rest
If 70% of your spend is m5 instances: commit $8,000/month as an EC2 Instance Savings Plan at 40% discount, remainder as Compute Savings Plan at 33%.
Annual savings: ($8,000 × 0.40 × 12) + ($4,500 × 0.33 × 12) = $38,400 + $17,820 = $56,220/year
Risk: Medium. If you migrate from m5 to m7g (Graviton), your EC2 Instance Savings Plan won't cover m7g instances — the Compute Savings Plan portion would, but you'd have unused EC2 Instance Savings Plan commitment.

The commitment trap: The biggest RI/Savings Plan mistake is over-committing. If you commit to $15/hr but only use $12/hr, you're paying for $3/hr of unused commitment. Always commit to your minimum baseline usage, not your average. Cover the floor with commitments, let the ceiling run on-demand. AWS Cost Explorer's Savings Plan recommendations use 7-day or 30-day lookback precisely to identify this baseline.

Payment Options: No Upfront vs. Partial vs. All Upfront

For both RIs and Savings Plans, you choose a payment option that affects the discount rate and your cash flow:

For a 1-year m5.2xlarge Standard RI in us-east-1 (Linux):

The incremental savings from All Upfront vs No Upfront on a single instance is ~$137/month. The ROI on the upfront payment depends on your cost of capital — for most companies, the incremental savings easily justify the upfront payment.

Managing Your RI/SP Portfolio

Purchasing commitments is the beginning, not the end. You need to actively manage your RI/SP portfolio:

Get a Personalized Commitment Strategy

Hero Savings analyzes your actual AWS usage patterns and generates a precise Savings Plan and RI purchasing strategy — showing you exactly which commitments to buy, in what amounts, to maximize savings without over-committing.

Optimize My Commitments →

Frequently Asked Questions

Can I use both Reserved Instances and Savings Plans?
Yes, and this is often the optimal strategy. AWS applies discounts in this order: (1) Standard RIs are applied first to matching usage, (2) EC2 Instance Savings Plans are applied next to the same instance family, (3) Compute Savings Plans are applied to any remaining usage. You might use Standard RIs for very stable, fixed instance types (e.g., a specific RDS configuration), EC2 Instance Savings Plans for stable EC2 families, and Compute Savings Plans for the remainder of your compute spend.
What happens if I purchase too many Reserved Instances?
If you have Standard RI commitments with no matching running instances, you're paying for the RI regardless. You have three options: (1) sell the RI on the AWS RI Marketplace (typically 80–95% recovery), (2) modify the RI to match instances you do run (if the RI allows modification), or (3) accept the loss and let it expire. This is the core risk of Standard RIs vs Savings Plans — Savings Plans automatically apply to any qualifying usage, so unused commitment is spread across all your compute rather than tied to a specific instance type.
Should I commit to 1-year or 3-year terms?
For most teams in 2026, 1-year terms are the right default for Compute Savings Plans. Cloud infrastructure evolves fast — your instance mix 3 years from now may look very different. The incremental savings from 3-year (66% vs 40% for Compute SPs) are significant, but only if your usage pattern is truly stable over that horizon. A practical approach: buy 1-year Compute Savings Plans as your base layer. Once you have 2+ years of stable usage data on specific services, consider 3-year commitments for those proven workloads.
Do Savings Plans cover RDS?
No. AWS Savings Plans (Compute and EC2 Instance) cover EC2, Fargate, and Lambda only. RDS has its own separate Reserved Instance product. If you're running significant RDS spend ($2,000+/month per database), RDS Reserved Instances should be on your FinOps radar — they save up to 69% on production databases with stable configurations.
How do I know if my current Savings Plans coverage is optimal?
In AWS Cost Explorer, go to Savings Plans → Coverage report. This shows what percentage of your eligible spend is covered by Savings Plans vs running on-demand. Coverage below 70% typically indicates significant untapped savings. Coverage above 95% may indicate you're over-committed. The sweet spot for most teams is 70–90% Savings Plan coverage on EC2, with the remaining on-demand for burst capacity. Hero Savings provides automated coverage analysis and purchase recommendations tailored to your specific usage patterns.