
CTO and SVP Engineering at Xcitium, running production workloads across AWS and Google Cloud. Profile
Last reviewed 26 September 2026
AWS EC2 Reserved Instances are a commitment-based discount: you agree to a specific instance configuration for one or three years, and pay less per hour than On-Demand in exchange. They were the main way to cut steady-state compute costs on AWS for most of the last decade.
They are also no longer what AWS leads with. The EC2 pricing page now presents three purchasing options — On-Demand, Savings Plans and Spot — and Reserved Instances are not among them. That does not mean they stopped working or that existing reservations are at risk. It does mean that if you are choosing a commitment model today, the default answer has changed.
This guide covers how AWS EC2 Reserved Instances work, what still makes them useful, and what to do if you hold some.
How AWS EC2 Reserved Instances work
A reservation is a billing construct, not a server. You do not launch a Reserved Instance — you launch a normal instance, and if it matches an active reservation, it bills at the reduced rate automatically. Nothing about the instance itself differs.
Three things define a reservation:
Term
One year or three years. Longer terms carry deeper discounts. Commitment is unconditional: the capacity is billed whether you use it or not, with no rollover for unused hours.
Payment option
All upfront, partial upfront, or no upfront. The discount is largest with all upfront and smallest with no upfront, with partial in between. All three commit you for the full term — the choice affects cash flow, not obligation.
Offering class
Standard reservations give the larger discount but lock the instance family. Convertible reservations give a smaller discount and allow you to exchange for a different family, OS or tenancy during the term — useful when you are reasonably sure about volume but not about shape.
Regional versus zonal AWS EC2 Reserved Instances
A reservation is scoped either to a region or to a specific availability zone, and the trade is flexibility against guarantee.
- Regional reservations apply to matching usage anywhere in the region and adjust automatically across availability zones. They do not reserve capacity.
- Zonal reservations are pinned to one availability zone and do reserve capacity there — a genuine guarantee that the instance will be available when you ask for it.
Most workloads want regional scope, because the flexibility is worth more than a capacity guarantee they will never need. Zonal makes sense when you must be certain capacity exists in a particular zone, typically for disaster recovery or regulated failover.
Why Savings Plans replaced AWS EC2 Reserved Instances as the default
Savings Plans commit to an amount of compute spend per hour rather than to a specific instance configuration. A Compute Savings Plan applies to EC2 usage regardless of instance family, size, operating system, tenancy or region, and extends to AWS Fargate and AWS Lambda as well. Terms are one or three years, with the same three payment options, and AWS states savings of up to 72% against On-Demand.
That difference is the whole story. A Reserved Instance stops paying off the moment you change instance family — which is exactly what you should be doing when AWS ships a newer generation with better price-performance. A Savings Plan follows you. The old model quietly penalised modernisation; the new one does not.
For the full current set, see EC2 purchasing options, and AWS’s own Savings Plans documentation for the mechanics.
AWS EC2 Reserved Instances versus Savings Plans
The two models solve the same problem — cheaper steady-state compute in exchange for commitment — but they commit you to different things.
| Reserved Instances | Savings Plans | |
|---|---|---|
| You commit to | A specific instance configuration | An amount of compute spend per hour |
| Terms | 1 or 3 years | 1 or 3 years |
| Payment options | All, partial or no upfront | All, partial or no upfront |
| Change instance family | Only with convertible, by exchange | Automatically covered |
| Change region | No | Covered by Compute Savings Plans |
| Covers Fargate and Lambda | No | Yes, with Compute Savings Plans |
| Reserves capacity | Zonal reservations only | No |
| Resellable | Standard only, via the Marketplace | No |
The row that decides it for most teams is capacity reservation. If you do not need a guarantee that a specific availability zone will have room for your instance — and most workloads genuinely do not — the flexibility of a Savings Plan is worth more than the marginally deeper discount a standard reservation might offer.
The row people overlook is Fargate and Lambda coverage. A team modernising away from EC2 toward containers and functions will watch a Reserved Instance become worthless as the migration succeeds, while a Compute Savings Plan follows the workload across all three.
When AWS EC2 Reserved Instances still make sense
They have not become useless. Two cases remain genuinely strong:
- You need guaranteed capacity in a specific availability zone. Savings Plans are purely a billing discount and reserve nothing. A zonal Reserved Instance is the mechanism that actually holds capacity for you.
- You hold reservations already. An active reservation continues to apply for its full term. There is no benefit to abandoning one you are using.
There is also a secondary market: standard Reserved Instances can be sold through the Reserved Instance Marketplace, which gives an exit for capacity you committed to and no longer need. Convertible reservations cannot be sold, only exchanged.
What to do if you hold AWS EC2 Reserved Instances
- Check coverage and utilisation in Cost Explorer. An under-used reservation is money already spent on capacity nobody is consuming, and nothing alerts you to it.
- Let them expire rather than renewing by default. When a term ends, a Savings Plan is usually the better replacement unless you specifically need zonal capacity.
- Exchange convertibles toward current generations. If you hold convertible reservations on older families, exchanging them toward newer ones captures the price-performance improvement you are otherwise locked out of.
- Do not stack commitments blindly. Savings Plans and Reserved Instances both apply to the same usage, and buying one while the other is active can leave you committed to more compute than you run.
- Right-size before you commit to anything. A three-year commitment on an oversized fleet locks in the waste. Fix the sizing first — see choosing EC2 instance types and the factors that affect EC2 pricing.
The general rule: commit to spend rather than to configuration, and commit only to the baseline you are confident you will run for the full term. Anything above that baseline is better served by Spot or On-Demand.