Reserved Instances vs Savings Plans: which one fits you
“Should we buy Reserved Instances or Savings Plans?” is one of the questions we hear most often once a company has been on AWS for a few months and starts looking at the bill more carefully. The short answer is that it depends on how predictable your architecture is — but it is worth understanding the real difference before committing to either one.
The real difference: what you are committing to
Both mechanisms give you a discount in exchange for a 1- or 3-year usage or spend commitment. The difference is what you commit to.
With Reserved Instances (RI) you commit to a specific instance: family (say, m5.large), region and, in some cases, availability zone. If your workload changes — you move from m5 to m6i, or shift workloads to another region — the reservation stops applying and you keep paying for a discount on something you no longer use.
With Savings Plans you commit to a spend amount per hour (for example, “$10/hour of compute”), regardless of which instance, family or region you use. The discount applies automatically to any eligible usage — EC2, Fargate and Lambda in the case of Compute Savings Plans.
The numbers: how much discount each one gives
- Reserved Instances: up to 72% off on-demand pricing, depending on the payment option (no upfront, partial upfront, or all upfront) and the term (1 or 3 years).
- Compute Savings Plans: up to 66% off, with the advantage of applying to any instance family, region, or even Fargate and Lambda.
- EC2 Instance Savings Plans: a middle ground — more restrictive than Compute Savings Plans (it applies to a specific instance family in a specific region), but with a higher discount, up to 72%.
In general: if you need the maximum possible discount and your workload is stable, RI or EC2 Instance Savings Plans get you there. If you need flexibility because your architecture is still changing, Compute Savings Plans is the safer choice, even if the maximum discount is somewhat lower.
When each one makes sense
Reserved Instances make sense when:
- You have already been running the same instance family in the same region for over a year, with no plans to move to a newer hardware generation.
- You run database workloads (RDS) that rarely change instance type.
- You want the maximum discount and are willing to trade flexibility for it.
Savings Plans make sense when:
- Your architecture is still evolving — you are migrating workloads, testing new instance families, or moving things to containers or serverless.
- You run a mix of EC2, Fargate and Lambda, and want the same commitment to cover all of it automatically.
- You have been on AWS for less than a year and do not have enough history yet to commit to a specific setup.
A practical case
A company with a stable base spend of $1,500/month on EC2, running the same instance family for 18 months, can buy EC2 Instance Savings Plans or 1-year Reserved Instances and capture roughly a 40-50% discount on that base spend — between $600 and $750 a month in savings, without changing anything in the architecture.
That same company, if it were in the middle of migrating from m5 instances to newer ones, or moving part of the workload to Fargate, would be better off with Compute Savings Plans: the spend commitment stays valid through the whole transition, without losing the discount when the instance type changes.
The most common mistake: buying before measuring
The mistake we see most often is buying Reserved Instances or Savings Plans without having measured real baseline usage for at least 2-3 months. Committing to a spend amount or a specific instance before you have clear visibility into your usage pattern usually ends in one of two ways: overcommitting (paying for capacity you do not use) or undercommitting (still paying on-demand for a meaningful share of load that was actually predictable).
Can I combine both?
Yes. In fact, it is common in mixed architectures: Reserved Instances for database workloads (RDS) that almost never change type, and Savings Plans for compute (EC2/Fargate/Lambda) that does evolve over time. They are not mutually exclusive — the key is identifying which part of your infrastructure is stable and which is still moving, and assigning the right mechanism to each.
What if I already bought the wrong one?
It happens more often than you would think. Reserved Instances can be sold on the AWS RI Marketplace (with some restrictions and a fee), and AWS lets you modify certain parameters of a reservation (availability zone, instance size within the same family) without losing the discount. Savings Plans, on the other hand, cannot be resold or cancelled — the spend commitment stays fixed for the term, which is exactly why measuring properly before buying matters so much.
Not sure which one you need?
At CloudTing we review your historical usage in Cost Explorer and tell you exactly what to buy, how much to cover, and what real savings to expect — before you commit to anything.
Request an AWS cost review →