---
title: "Savings Plans and Reserved Instances: A Commitment Strategy That Does Not Backfire"
description: "Commitments are the biggest single discount on AWS and the easiest way to lock in waste for three years. How we size, sequence and monitor Savings Plans and Reserved Instances for EU enterprises."
doc_version: 1.0.0
last_updated: 2026-10-02
date_published: 2026-09-09
canonical: https://ccx.hu/blog/aws-savings-plans-reserved-instances-commitment-strategy
---

# Savings Plans and Reserved Instances: A Commitment Strategy That Does Not Backfire

> Commitments are the biggest single discount on AWS and the easiest way to lock in waste for three years. How we size, sequence and monitor Savings Plans and Reserved Instances for EU enterprises.

![Savings Plans and Reserved Instances: A Commitment Strategy That Does Not Backfire](https://s3.us-east-1.amazonaws.com/ccx.hu/images/01-savings-plans-and-reserved-instances.png)

Commitments are the biggest single discount on AWS, and the easiest way to lock in waste for three years. We regularly see estates with high Savings Plan coverage and a bill that is still too large: the commitment was sized against oversized instances, bought in one block, and never revisited. The discount was real; it was applied to capacity nobody needed. This post is how we size, sequence and monitor commitments so they keep paying back.

## Two numbers, not one

Every commitment conversation should start with two numbers, tracked separately in Cost Explorer:

- **Coverage** - the share of eligible usage that ran at a discounted rate. Low coverage means you are paying on-demand prices for steady workloads.
- **Utilisation** - the share of what you committed to that was actually used. Anything below 100% is money paid for nothing.

They pull in opposite directions. Buying more raises coverage and risks utilisation; buying less protects utilisation and leaves coverage on the table. A dashboard that reports only coverage rewards overcommitting, which is how teams end up with 90% coverage and an unused commitment line on every invoice. High coverage on the wrong commitment is a loss, not a win.

## Which instrument covers what

| Instrument                      | Applies to                                                       | Flexibility                                  |
| ------------------------------- | ---------------------------------------------------------------- | -------------------------------------------- |
| Compute Savings Plan            | EC2 (any family, size, region, OS, tenancy), Fargate, Lambda     | Highest; follows you through migrations      |
| EC2 Instance Savings Plan       | One instance family in one region, any size, OS or tenancy       | Deeper discount, locked to family and region |
| SageMaker Savings Plan          | SageMaker AI instance usage                                      | Any family, size or region within SageMaker  |
| Reserved Instances              | RDS, ElastiCache, OpenSearch, Redshift, and EC2                  | Engine, family and region specific           |
| Database Savings Plans          | Managed database services, per AWS's current list of engines     | Newer instrument; check coverage before buying |

A few practical rules fall out of that table:

- **Compute Savings Plans are the default.** They give up some discount compared with EC2 Instance Savings Plans, and in exchange cover instance-family changes, Graviton moves, region changes and the shift from EC2 to containers or Lambda. For estates still modernising, that flexibility is usually worth more than the extra points of discount.
- **EC2 Instance Savings Plans fit only workloads you are confident will stay on one family in one region** for the full term.
- **Databases need their own plan.** Compute Savings Plans do not touch RDS or ElastiCache. AWS introduced Database Savings Plans at the end of 2025; compare their current engine coverage and discount against Reserved Instances for your specific engines before deciding.

## Sequence: shrink first, commit second

The order matters more than the instrument.

1. **Remove idle and rightsize first.** Every unattached volume, forgotten environment and oversized instance you commit against becomes a three-year obligation. Rightsizing after a purchase does not reduce the commitment; it reduces utilisation.
2. **Commit to the floor, not the average.** Look at hourly eligible on-demand spend over the last 30 to 60 days, after rightsizing. Commit near the lower end of that range, the level you are confident will still be running every hour for the term. Peaks stay on-demand or move to Spot.
3. **Flexible instruments first, specific ones second.** A Compute Savings Plan for the stable compute floor, then EC2 Instance Savings Plans only for families you are sure about, then database commitments engine by engine.

### A worked example (illustrative numbers)

An estate's eligible compute spend, after rightsizing, runs between $38 and $61 an hour on demand over the last 60 days, with a clear daily pattern and a floor of about $40.

- Committing against the average (about $50 an hour on demand) means the commitment goes partly unused every night and weekend.
- Committing against the $40 floor, and applying a blended Compute Savings Plan discount of roughly 30% for this mix, gives an hourly commitment of about $28. That is fully used every hour, and the peaks above the floor are still billed on demand.

Savings Plan commitments are expressed as dollars per hour of *discounted* spend, which is why the commitment is lower than the on-demand floor it covers. The real discount depends on your instance mix, region, term and payment option; use the AWS recommendation and pricing tools for your own numbers.

## Ladder purchases instead of buying in one block

A single large purchase creates a single large expiry date and a single large bet on today's forecast. Laddering spreads both:

- Buy in **monthly or quarterly tranches**, each covering part of the target.
- Each tranche is sized against the floor **as it looks at the time of purchase**, so later tranches reflect rightsizing and migrations that have already happened.
- Expiries are staggered, so a change in strategy (a Graviton migration, a move to containers, a workload retired) is absorbed by simply not renewing the next tranche.

Laddering costs a little coverage in the first months. It removes the scenario where a three-year commitment bought in one afternoon outlives the architecture it was sized for.

## Term and payment option

Three-year terms and all-upfront payment give the deepest discounts, and both increase the cost of being wrong. Our defaults:

- **One-year, no upfront** for anything still changing: estates mid-migration, workloads with an uncertain roadmap, or the first ladder tranches while the floor is still being established.
- **Three-year** only for the part of the floor you would bet on in a board meeting: core platforms, regulated systems with long lifetimes.
- **Upfront payment** is a treasury decision as much as an engineering one. Bring finance in; the extra discount is effectively a return on cash paid early, and they will have a view on whether that is a good use of it.

Remember what cannot be undone. Savings Plans cannot be resold, and AWS allows returning a Savings Plan only within a short window after purchase and under specific limits. Standard EC2 Reserved Instances can be listed on the Reserved Instance Marketplace, with conditions; most other commitments are simply yours for the term.

## Using the AWS recommendations without trusting them blindly

Cost Explorer produces Savings Plan and Reserved Instance purchase recommendations based on a 7, 30 or 60 day lookback. They are a useful starting point:

```bash
aws ce get-savings-plans-purchase-recommendation \
  --savings-plans-type COMPUTE_SP \
  --term-in-years ONE_YEAR \
  --payment-option NO_UPFRONT \
  --lookback-period-in-days SIXTY_DAYS
```

What the recommendation does not know:

- **Your roadmap.** It assumes the last 60 days repeat for the whole term. It does not know the batch platform is being retired in March.
- **Your rightsizing backlog.** It sizes against current usage, including the waste you are about to remove.
- **Your organisation's other purchases.** Run it at the payer account level, with discount sharing understood, rather than per account.

Treat it as an upper bound, then cut it back to the floor you are confident in.

## Organisation-level sharing

By default, Savings Plan and Reserved Instance discounts are shared across all accounts in an AWS Organization's consolidated bill. That raises utilisation, because a commitment bought for one account can absorb usage in another. It also means the account that bought the commitment does not necessarily get the benefit.

Decide deliberately:

- **Keep sharing on** and buy centrally from the management or a dedicated billing account, if commitments are a central finance function.
- **Turn sharing off for specific accounts** if a business unit buys and owns its own commitments and needs its showback to reflect that.

Either way, report amortised cost to teams, so commitment benefits and fees land on the workloads that consume them. Without that, showback is skewed by whoever happens to run on covered capacity.

## Monitoring after the purchase

Commitments drift as soon as they are bought. Four controls keep them honest:

- **AWS Budgets for Savings Plan and Reserved Instance utilisation and coverage**, alerting when utilisation drops below a threshold, typically the first sign that a workload moved or shrank.
- **A monthly utilisation and coverage check** alongside the cost review:

  ```bash
  aws ce get-savings-plans-utilization \
    --time-period Start=2026-09-01,End=2026-10-01
  ```

- **Expiry tracking.** Lapsed commitments quietly return workloads to on-demand rates. Put each tranche's expiry date in the same calendar as contract renewals.
- **A re-run of the floor analysis before every ladder tranche**, so the next purchase reflects the estate as it is, not as it was.

## What this looks like at ccX

Commitment strategy is the first of the six levers in our [two-week AWS Cost Optimization Review](https://ccx.hu/aws-cost-optimization), and it is deliberately sequenced after rightsizing and idle clean-up so you never commit against waste. We model Savings Plans and Reserved Instances against your actual hourly usage curve, recommend a laddered purchase plan with coverage and utilisation targets, and set up the budgets and alerts that keep it on track. We currently run commitment strategy in production across an EU institution's multi-account estate. Fixed fee, read-only access, never a percentage of savings.

If your Savings Plans were bought once and have not been looked at since, [book a 30-minute scoping call](https://ccx.hu/book).

## Where to go from here

- [AWS Cost Optimization Review](https://ccx.hu/aws-cost-optimization) - the full two-week engagement.
- [AWS cost allocation that survives real teams](https://ccx.hu/blog/aws-cost-allocation-tagging-and-showback) - allocation first, so every commitment has an owner.
- [Contact us](https://ccx.hu/contact)

> *Discount levels, eligible services and return terms for Savings Plans and Reserved Instances change over time. Verify them against current AWS pricing and documentation before purchasing.*

---

## Sitemap

- [Read this post on the web](https://ccx.hu/blog/aws-savings-plans-reserved-instances-commitment-strategy)
- [All blog posts](https://ccx.hu/blog)
- [Full site map](https://ccx.hu/sitemap.md)
- [Home](https://ccx.hu/)
- [Services](https://ccx.hu/services)
- [EU AI Act Compliance](https://ccx.hu/eu-ai-act)
- [AWS GenAI Production Readiness](https://ccx.hu/aws-genai-review)
- [Glossary](https://ccx.hu/glossary)
- [Contact](https://ccx.hu/contact)
