AWS Savings Plans calculator: Savings Plans vs Reserved Instances

Both give you a lower EC2 price in exchange for a commitment of one year (365 days) or three years (1,095 days). Neither is a server. A Reserved Instance is a billing discount applied to matching on-demand usage, and a Savings Plan applies its rates to eligible usage automatically.

Savings Plans commit you to spend a fixed amount in dollars per hour for the term. Usage is charged at the plan's rates until the hour's commitment is used up; anything above it is charged at on-demand rates, and commitment left over in an hour cannot be carried over. A Compute Savings Plan applies to EC2 regardless of instance family, size, Region, operating system or tenancy, and also to Fargate and Lambda. An EC2 Instance Savings Plan is tied to one instance family in one Region, applies to any size, operating system and tenancy within it, and has a lower rate.

Reserved Instances are priced by instance type, Region, tenancy and platform. A Standard Reserved Instance has the larger discount; you can modify its Availability Zone, scope and size within the same family and generation, and sell it on the Reserved Instance Marketplace, but you cannot exchange it. A Convertible Reserved Instance has a smaller discount and can be exchanged for another Convertible Reserved Instance with a different instance family, operating system or tenancy, as long as the new one is of equal or higher value. It cannot be sold on the Marketplace. A regional Reserved Instance applies in any Availability Zone of its Region and, for Linux with default tenancy, to any size in its family, except in some GPU, accelerator and HPC families, including G4dn, G5, Inf2 and hpc7a, where it covers only its own size; only a zonal Reserved Instance reserves capacity.

Both come with three payment options. No Upfront bills the whole commitment over the term. Partial Upfront takes part of it at the start (at least half, for a Savings Plan) and bills the rest over the term. All Upfront takes everything at the start. A larger upfront payment generally lowers the total. A Reserved Instance bills every hour of the term whether or not you use it, and when it expires your instances go back to on-demand rates. AWS's EC2 documentation recommends Savings Plans over Reserved Instances.

10 m7i.large instances in US East (N. Virginia), Linux, running every hour of a 1-year term, bought with No Upfront. Each line is an alternative for the same ten instances.

On-demand = $0.1008 × 8,760 h × 10 = $8,830
Compute Savings Plan = $0.07406 × 8,760 h × 10 = $6,488 (27% less, pays off above 73.5% use)
EC2 Instance Savings Plan = $0.06668 × 8,760 h × 10 = $5,841 (34% less, pays off above 66.2% use)
Standard Reserved Instance = $0.06668 × 8,760 h × 10 = $5,841 (34% less, pays off above 66.2% use)
Convertible Reserved Instance = $0.07406 × 8,760 h × 10 = $6,488 (27% less, pays off above 73.5% use)

The cheapest are the EC2 Instance Savings Plan and the Standard Reserved Instance, at $5,841 each: $2,989 less than $8,830 on-demand for the year.

On a 3-year All Upfront term the same ten instances cost $10,459 with the EC2 Instance Savings Plan, paid on the first day, against $26,490 on-demand over three years: 61% less, if you still need all ten in year three.

A commitment is paid whether or not you use it. The break-even share in the table is the point where a commitment starts to beat on-demand: for the EC2 Instance Savings Plan and the Standard Reserved Instance above, the instances must run more than 66.2% of the hours in the term.

If those ten instances turn out to be needed only half the time, because the workload shrinks, moves to another family or Region, or is switched off at night, on-demand would have cost $4,415 for the year. The commitment still costs $5,841: $1,426 more than not committing at all.

The narrower the commitment, the more of it can go unused. A Compute Savings Plan follows your usage to other instance families, Regions, Fargate and Lambda; an EC2 Instance Savings Plan only to other sizes, operating systems and tenancies in the same family and Region; a Reserved Instance only to instances that match it, or, for a regional Linux Reserved Instance outside the families above, to other sizes in its family. A 3-year term buys a larger discount by making the same bet for three times as long.

Spot Instances use spare EC2 capacity. The Spot price for each instance type in each Availability Zone is set by EC2 and adjusted gradually with long-term supply and demand. It is not in the AWS Price List, so this page shows a sample: at 2026-09-19 19:36 UTC, m7i.large in US East (N. Virginia) was $0.0384 to $0.0526 an hour across 5 Availability Zones, against $0.1008 on-demand (48% to 62% less).

The trade-off is interruption. EC2 can stop or terminate a Spot Instance when it needs the capacity back, with a two-minute warning. That suits work that can be restarted or spread out, such as batch jobs, data analysis and background processing. Savings Plans do not cover Spot Instances, and Spot spend does not count toward a Compute Savings Plan commitment. Availability Zone names map to different zones in each AWS account, so the sample is shown as a range, not per zone.

• Rightsize before you commit. A commitment locks in today's spend for one or three years, so commit to what you will still run after removing idle and oversized instances.
• Commit to the floor, not the peak. Cover the usage that runs every hour with a commitment, and leave the peaks on on-demand or Spot.
• Pay for flexibility you will use. If you may change instance family or Region, or move to Fargate or Lambda, a Compute Savings Plan keeps the discount. If the family and Region are settled, an EC2 Instance Savings Plan or a Standard Reserved Instance costs less.
• Buy in steps. Several smaller purchases at different dates leave you room to adjust as terms end, instead of one large commitment that ends all at once.
• Start from your own usage. AWS Cost Explorer shows a recommended Savings Plans commitment for your account, next to your historical usage.