AWS Savings Plans: What They Are and When to Use Them

Definition

AWS Savings Plans are a pricing model in which you commit to a consistent amount of compute usage (measured in $/hour) for a 1-year or 3-year term in exchange for a significant discount — up to 72% off on-demand rates. Savings Plans replaced most use cases for EC2 Reserved Instances when they launched in 2019 and are now the default recommendation for almost every AWS cost-optimization effort.

How It Works

You commit to a fixed $/hour of compute spend over the term — for example, "I will use at least $5/hour of compute for the next 3 years." AWS discounts the first $5/hour of matching compute on every hour in the Region (or globally, depending on plan type) at the agreed rate. Any usage above that runs at on-demand prices.

Billing is calculated hourly, so a spiky day balances out against a quiet day — as long as the average hourly spend is at least your commitment, you get the discounted rate on every eligible dollar.

Commitment payment options:

• No Upfront — pay monthly, modest discount.
• Partial Upfront — pay half at the start and half over the term; larger discount.
• All Upfront — pay the entire commitment up front; deepest discount.

The Three Types of Savings Plans

1. Compute Savings Plans

• Discount: up to 66% off On-Demand.
• Flexibility: apply to EC2 (any instance family, any Region, any OS, any tenancy), AWS Fargate, AWS Lambda, and EC2-equivalent SageMaker Inference.
• When: the default for most organizations. Maximum flexibility lets teams change compute platforms (migrate EC2 to Fargate, rearchitect to Lambda) without losing the discount.

2. EC2 Instance Savings Plans

• Discount: up to 72% off On-Demand.
• Flexibility: locked to a specific instance family in a specific Region (e.g., "m7g in us-east-1"). Within that family/Region you can still change instance size, OS, and tenancy.
• When: you have stable, predictable EC2 usage of a specific family in a specific Region for years — for example, the core fleet of a mature production workload. The deeper discount is the reward for less flexibility.

3. SageMaker Savings Plans

• Discount: up to 64% off On-Demand.
• Flexibility: applies to SageMaker training, inference, and notebook instances across any Region, any instance family, any component.
• When: you have material, steady SageMaker usage and want to hedge against ML workload growth across instance families.

Key Features and Limits

• Term: 1 year or 3 years. 3-year terms offer deeper discounts.
• Payment options: No Upfront, Partial Upfront, All Upfront.
• Billing account sharing: Savings Plans apply across linked accounts in an AWS Organization automatically (unless opted out).
• Cannot be cancelled or modified — you own the commitment for the term. You can only let it expire and replace it.
• Convertible: Savings Plans are not convertible between types (e.g., you cannot convert a Compute Savings Plan to an EC2 Instance Savings Plan).
• Coverage reports: Cost Explorer shows Savings Plan utilization and coverage, with recommendations based on your historical usage.
• Queueable: you can queue a plan to start up to 3 years in the future, useful for staggered expiries and to lock in current prices.

Common Use Cases

1. Any production EC2 or Fargate workload you run 24/7 year-round — nearly always worth a Compute Savings Plan for the baseline.
2. Lambda-heavy serverless architectures — Compute Savings Plans apply to Lambda duration GB-seconds (though not per-request cost).
3. Fleet standardization — if you have settled on one instance family in one Region, an EC2 Instance Savings Plan for that family gives the deepest discount.
4. ML platforms with steady SageMaker usage — SageMaker Savings Plans.
5. Hedging future growth — queue a second plan today to start when your first expires, locking in current rates.

Pricing Model

• Savings Plans charges follow the payment option you selected: one large up-front charge + remaining spread over the term, or no upfront with monthly billing.
• Usage that matches your plan is billed at the discounted rate automatically; usage above is billed On-Demand.
• Unused commitment is still charged — if you commit to $5/hour and use $3/hour, you still pay $5/hour.
• Spot Instances, EC2 Dedicated Hosts (by default), and some specialty pricing are not eligible. Always check the coverage reports.

In Cost Explorer, the Savings Plans Coverage and Utilization reports tell you what percentage of eligible usage is covered and what percentage of your commitment is being consumed.

Pros and Cons

Pros

• Largest lever for AWS cost reduction for steady workloads — 40–72% off.
• Compute Savings Plans are flexible across EC2, Fargate, and Lambda.
• Automatic application across accounts in an Organization.
• Simpler than Reserved Instances — you commit to $/hour, not specific instances.
• Can be combined with Spot Instances (Savings Plans cover baseline; Spot covers burst capacity).

Cons

• Commitment is irrevocable — if your workload shrinks or shifts off AWS, you still pay.
• Not convertible between plan types.
• Doesn't cover every service (no Redshift, RDS, ElastiCache, Lightsail — those have their own Reserved pricing).
• Understanding which workloads to commit to takes analysis; Cost Explorer recommendations help but should be reviewed.

Comparison with Alternatives

| | Savings Plans | Reserved Instances | Spot Instances | On-Demand | | --- | --- | --- | --- | --- | | Commitment | $/hour | Specific instance config | None | None | | Term | 1 or 3 years | 1 or 3 years | Per-hour / per-second | Per-second | | Max discount | 66% (Compute) / 72% (EC2 Instance) | 72% (Standard) | Up to 90% | 0% | | Flexibility | High (Compute) / Medium (EC2 Instance) | Low (Standard) / Medium (Convertible) | None — interruptible | Full | | Covers | EC2, Fargate, Lambda, SageMaker | EC2 only (similar for RDS, Redshift, ElastiCache under those services' RI) | EC2 | All | | Ideal for | Steady baseline workloads | Legacy commitments, RDS/Redshift | Fault-tolerant batch, CI/CD | Unpredictable, short-term |

Rule of thumb: for every long-term workload, layer these in order:

1. Savings Plans (Compute, or EC2 Instance for stable fleets) cover the predictable baseline.
2. Spot Instances cover burst or fault-tolerant capacity.
3. On-Demand covers everything else (development, unpredictable spikes).

Exam Relevance

• Cloud Practitioner (CLF-C02) — know Savings Plans provide the largest discount for committed usage, 1- or 3-year terms.
• Solutions Architect Associate (SAA-C03) — frequent scenario: "the company has steady compute usage and wants to reduce cost" → the answer is almost always Compute Savings Plans.
• Solutions Architect Professional (SAP-C02) — portfolio optimization: combining Savings Plans with Spot Instances, commitment strategies across accounts in an Organization.
• DevOps Professional (DOP-C02) — automating Savings Plan purchase via CloudWatch / Cost Explorer APIs.

Classic exam trap: Savings Plans are not cancellable. If the scenario implies that the workload might go away in six months, the answer is On-Demand or Spot — not a 1-year Savings Plan.

Frequently Asked Questions

Q: What's the difference between Compute and EC2 Instance Savings Plans?

A: Compute Savings Plans apply to EC2, Fargate, Lambda, and equivalent SageMaker Inference — across any Region, instance family, OS, and tenancy. Maximum flexibility; up to 66% off. EC2 Instance Savings Plans apply only to a specific instance family in a specific Region (e.g., m7g in us-east-1), but within that scope, OS/tenancy/size can change freely. Less flexible; up to 72% off. Pick Compute for most workloads; pick EC2 Instance for a stable, large fleet locked to one family.

Q: Can I cancel a Savings Plan if my workload goes away?

A: No. Savings Plans are a firm commitment for the full 1-year or 3-year term and cannot be cancelled, modified, or sold on a marketplace. If a workload ends, the commitment continues — you either replace the usage with other eligible workloads or absorb the cost. This is why conservative sizing (e.g., covering 60–80% of your baseline rather than 100%) is usually the right strategy.

Q: Do Savings Plans work with Spot Instances?

A: No — Spot Instances are already deeply discounted and are not eligible for Savings Plans coverage. The common architecture is to layer them: a Compute Savings Plan covers the on-demand baseline you always need, and Spot Instances handle burst capacity or fault-tolerant workloads. This hybrid can drop total compute cost by 60–80% compared to 100% On-Demand.

---

This article reflects AWS features and pricing as of 2026. AWS services evolve rapidly — always verify against the official AWS Savings Plans documentation before making production decisions.