AWS Compute Savings Plans reduce the price of eligible Amazon EC2, AWS Fargate and AWS Lambda usage in exchange for a fixed spending commitment, measured in dollars per hour, for one or three years. The discount follows eligible usage as your architecture changes. You still owe the commitment during hours when you use less. AWS Savings Plans overview.
AWS advertises savings of up to 66% for Compute Savings Plans. Your actual savings depend on the applicable rates and how much commitment your workloads consume. A useful purchase decision starts with the compute you expect to keep running, including quieter hours. AWS Compute Savings Plans pricing.
Account ownership and workload mix can change the value of a fully utilized plan. The examples below show why the same $9/hour commitment can save $1 or $6 per hour, and why equal daily usage totals can produce different bills.
What do AWS Compute Savings Plans cover?
A Compute Savings Plan can apply across these services, subject to eligibility for each charge:
| Service | Eligible usage | What to check separately |
|---|---|---|
| Amazon EC2 | Eligible instance usage across families, sizes, Availability Zones, Regions, operating systems and tenancy | Separately billed storage, networking and software charges |
| AWS Fargate | Eligible vCPU and memory usage | Operating system license fees are excluded |
| AWS Lambda | Eligible compute charges, including function duration and Provisioned Concurrency | Requests receive no rate discount but can still consume commitment |
This flexibility can support a move between EC2 families or Regions, or from EC2 to Fargate, while eligible demand continues. Fargate discounts apply to specific CPU and memory dimensions, so review the AWS pricing details and Lambda Savings Plans eligibility before counting your entire service bill as eligible.
Compute Savings Plans do not cover Spot usage or usage already covered by EC2 Reserved Instances. They also do not reserve capacity. AWS Savings Plans and Reserved Instances.
Database and SageMaker AI usage have separate Savings Plan products. An EKS cluster can have eligible underlying EC2 or Fargate compute, while the EKS service charge remains separate. AWS service and plan distinctions.
For a detailed service breakdown, read our Compute Savings Plan coverage guide for Lambda and Fargate.
How AWS Compute Savings Plans work
You choose an hourly dollar commitment. AWS applies Savings Plans rates to eligible usage until the commitment is consumed. Remaining usage is billed On-Demand unless another applicable discount covers it.
Each hour is evaluated separately. A busy afternoon cannot recover commitment left unused overnight. AWS application and hourly billing rules.
The commitment is measured at Savings Plans rates. A $14 hourly commitment with an assumed uniform 30% discount can cover usage that would cost $20 at On-Demand prices.
AWS Compute Savings Plan example
Assume the same 30% discount applies to all eligible usage, existing commitments have already been accounted for, and no other usage can absorb unused commitment.
| Eligible usage at On-Demand prices | Commitment consumed | Extra On-Demand cost | Total effective hourly cost | Savings or loss |
|---|---|---|---|---|
| $20 | $14 | $0 | $14 | $6 saved |
| $40 | $14 | $20 | $34 | $6 saved |
| $10 | $7 | $0 | $14 | $4 loss |
During the quiet $10 hour, only $7 of commitment is used. The other $7 remains payable, making the hour more expensive than On-Demand.
These are illustrative calculations, not AWS quotes. Total effective cost includes the hourly allocation of any upfront payment and excludes additional fees.
How AWS decides which usage gets the Savings Plan discount
AWS applies EC2 Reserved Instances before Savings Plans, and EC2 Instance Savings Plans before Compute Savings Plans.
For the eligible usage that remains, the purchasing account gets priority before benefits are shared. Within the applicable allocation scope, AWS uses the highest savings percentage first. Equal percentages are ordered by the lower Savings Plans rate. AWS discount application rules.
“Highest savings percentage” can favor a smaller instance over a more expensive one.
At a 40% discount, each $1 of consumed commitment covers approximately $1.67 of On-Demand-equivalent usage and saves $0.67. At a 20% discount, it covers $1.25 and saves $0.25, assuming sufficient eligible usage at each rate.
The calculation is:
On-Demand-equivalent usage covered = commitment consumed ÷ (1 − discount rate)
The purchasing account’s priority means AWS does not simply search for the highest discount anywhere in your organization. AWS describes this purchasing-account preference as account affinity. AWS guidance on account ownership and recommendations.
The same fully utilized plan can save $1 or $6 per hour
Consider two accounts:
- Account A has $10/hour of eligible Lambda usage with an illustrative 10% discount.
- Account B has $15/hour of eligible EC2 usage with an illustrative 40% discount.
Either workload consumes exactly $9/hour at its Savings Plans rate.
Compare purchasing the same $9/hour Compute Savings Plan in Account A with purchasing it in a dedicated account that has no eligible usage.
Assume organization-wide sharing is enabled, both workload accounts can receive benefits, and there are no other commitments or fees.
| Purchase location | Usage covered first | Plan utilization | Total hourly cost | Savings versus $25 On-Demand |
|---|---|---|---|---|
| Lambda Account A | Account A’s $10 of Lambda usage | 100% | $9 commitment + $15 EC2 = $24 | $1/hour |
| Empty purchasing account | Account B’s $15 of EC2 usage | 100% | $9 commitment + $10 Lambda = $19 | $6/hour |
The difference is $5/hour, or $3,600 over an illustrative 720-hour month, with identical workloads and 100% utilization in both cases.
Utilization measures commitment consumption. It does not show whether the purchasing arrangement captured the strongest available rate savings.
Compare purchase locations before committing, alongside chargeback and governance requirements. This example compares new purchases; it does not assume an existing plan can be transferred between accounts. The service discounts are hypothetical, not AWS quotes.
Sharing groups can change which demand is available
With Prioritized Group Sharing, unused benefits can move beyond the group after its eligible demand is satisfied.
Restricted Group Sharing keeps benefits inside the defined group, even when commitment remains unused. Both modes preserve the owner’s priority.
The management account cannot belong to a sharing group. An empty purchasing account helps only when the sharing configuration lets it reach the intended workloads. AWS sharing modes and account-group requirements.
The discount can follow a migration while the savings shrink
Compute Savings Plans remain flexible across eligible services. However, the dollar commitment can cover a different amount of On-Demand-equivalent usage when the applicable discount changes.
Hold eligible On-Demand-equivalent spending at $100/hour and the commitment at $50/hour. Compare an illustrative 50% discount before a migration with a 20% discount afterward.
| Measure | Before migration | After migration |
|---|---|---|
| Eligible usage at On-Demand prices | $100/hour | $100/hour |
| Assumed discount | 50% | 20% |
| On-Demand-equivalent usage covered by $50 | $100/hour | $62.50/hour |
| Total hourly cost | $50 | $50 + $37.50 uncovered = $87.50 |
| Commitment utilization | 100% | 100% |
| Hourly savings | $50 | $12.50 |
The plan remains fully utilized, but its savings fall because the replacement usage receives a smaller discount.
This calculation tests sensitivity to billing rates. It assumes a uniform rate within each scenario and no other discounts or fees; it does not compare the performance or total application cost of different services.
Reprice the post-migration usage at the plan’s applicable rates before deciding whether the existing commitment still fits. AWS publishes separate rates for each type of compute usage. AWS Savings Plans rates.
Can intermittent Fargate workloads share one Savings Plan?
Eligible workloads can consume the same plan’s hourly commitment when they are within its sharing scope. What matters is their combined demand in each hour, including when their schedules overlap.
Suppose two tasks each run for 12 hours per day and each costs $1/hour On-Demand while running.
Assume an illustrative 30% discount and a $0.70/hour commitment, with no other eligible usage, commitments or fees.
| Task schedules | Total task-hours per day | Daily plan plus On-Demand cost | Daily savings versus $24 On-Demand |
|---|---|---|---|
| One runs during each half of the day | 24 | $0.70 × 24 = $16.80 | $7.20 saved |
| Both run during the same 12 hours | 24 | ($0.70 × 24) + ($1 × 12 uncovered) = $28.80 | $4.80 loss |
The daily usage total is identical, but the overlapping schedule leaves half the commitment unused and part of the busy period On-Demand.
Evaluate aggregate hourly demand inside the actual sharing pool before concluding that an intermittent workload cannot benefit, or that a monthly usage total supports a purchase. AWS hourly application rules.
Lambda requests can consume commitment without receiving a discount
AWS distinguishes receiving a discount from consuming commitment.
Its Lambda launch guidance explains that request spend can draw down Savings Plans commitment even though requests receive no rate discount. A dollar consumed by a zero-discount request charge produces no rate saving on that charge. AWS explanation of Lambda request treatment.
For a Lambda-heavy account, check which billing dimensions consume the plan and the savings each produces. Higher utilization can reflect a shift toward lower-discount or zero-discount charges.
Our Lambda and Fargate coverage guide explains the distinctions between charges.
How to check which usage actually received the benefit
AWS Cost and Usage Reports can show an On-Demand-valued usage line even when Savings Plans covered it. Check the matching billing records before treating that line as an additional charge.
| Billing item or field | What it tells you |
|---|---|
SavingsPlanCoveredUsage | Usage that received the benefit. Its unblended cost shows the On-Demand equivalent. |
SavingsPlanNegation | The offset to that On-Demand-valued covered usage. A grouped offset can match several usage rows. |
savingsPlan/SavingsPlanEffectiveCost | The upfront and recurring commitment cost allocated to a covered usage line. |
Use the plan identifier and usage account to investigate allocation. Reconcile unused commitment separately when measuring the whole plan’s economics.
Adding allocated effective cost to the full plan fees would count the allocated portion twice. Keep cash-payment and amortized-cost views distinct. AWS Savings Plans billing records, AWS Savings Plans column definitions.
AWS Compute Savings Plans pricing and payment options
There is no single Compute Savings Plan discount for every workload. Compare the actual Savings Plans and On-Demand rates for the services and configurations you use, under each term and payment option.
For example, if a usage unit costs $1.00 On-Demand and $0.70 under your proposed plan, its rate discount is 30%. Your overall savings will also reflect unused commitment and charges that remain On-Demand.
The advertised maximum should not be applied to your total AWS bill.
Compute Savings Plans offer these payment options:
| Payment option | How you pay |
|---|---|
| No Upfront | No initial upfront payment; the commitment is billed monthly. |
| Partial Upfront | Pay part of the commitment at purchase and the remainder over the term. |
| All Upfront | Pay the entire commitment at purchase. AWS offers its lowest Savings Plans prices with this option. |
No Upfront still commits you for the selected one-year or three-year term. Compare the rate benefit with your cash requirements and confidence in future demand. AWS Savings Plans payment options.
Buying a larger hourly commitment does not improve the unit price for the same offering. It increases the usage you can cover and the amount you owe. AWS pricing comparison.
Compute Savings Plans vs EC2 Instance Savings Plans
| Decision factor | Compute Savings Plans | EC2 Instance Savings Plans |
|---|---|---|
| Eligible services | EC2, Fargate and Lambda | EC2 |
| EC2 family and Region | Flexible across eligible families and Regions | One selected family in one Region |
| Advertised maximum savings | Up to 66% | Up to 72% |
| Useful starting point | Mixed services or expected configuration changes | Durable EC2 demand within one family and Region |
The maximum percentages are product ceilings, not a guaranteed six-percentage-point difference for your instances. Compare actual rates against the architecture you expect to operate throughout the term. AWS plan comparison.
Use our Compute vs EC2 Instance Savings Plans comparison for the detailed tradeoffs.
For the broader product family, see our AWS Savings Plans guide.
How to choose your hourly commitment
Start with demand that will remain after optimization
Ask engineering which workloads will be resized, shut down or migrated during the proposed term. Build the commitment around the eligible demand expected to remain.
A historical bill can include temporary projects and resources already scheduled for removal.
Account for existing commitments and sharing
Use the remaining eligible usage after existing benefits to size a new purchase. Apply the account-ownership and sharing rules above when determining which workloads can consume the commitment.
Check inventory, expiration dates and queued purchases. Compute and EC2 Instance recommendations use the same underlying usage, so buying both recommendations in full can double-count the opportunity. AWS recommendation calculations.
Model quiet hours and future reductions
Compare hourly patterns across nights, weekends and seasonal changes. A monthly average hides how often demand falls below a proposed commitment.
For each candidate amount:
- Calculate the full commitment cost plus uncovered On-Demand charges.
- Repeat the calculation for planned rightsizing and a realistic reduction in demand.
- Compare each scenario with the cost of that same reduced workload without the new plan.
- Include management fees and applicable contracted rates when evaluating net savings.
A smaller commitment can preserve savings during weaker hours. A larger one can save more when demand persists. Choose between them using the expected full-term economics and the financial downside you can accept.
Use AWS recommendations as historical evidence
AWS recommendations use past usage, not a forecast of future demand, and do not account for queued or scheduled purchases. Refresh them after inventory changes and adjust for events outside the lookback window. How AWS calculates recommendations.
The Savings Plans Purchase Analyzer lets you compare recommended commitments, target coverage and custom hourly amounts.
Use those results alongside engineering’s workload plans before approving a purchase.
How to buy an AWS Compute Savings Plan
- Open AWS Billing and Cost Management and review your Savings Plans inventory and recommendations.
- In Purchase Analyzer, select Compute Savings Plans and the appropriate payer or linked-account analysis level.
- Compare one-year and three-year terms, payment options and candidate hourly commitments.
- Review the estimated cost, utilization and coverage, then account for upcoming workload changes and purchases.
- Complete the Savings Plans purchase flow after confirming the commitment, term, payment option and start date.
The AWS Purchase Analyzer instructions describe the analysis settings.
Record the purchase assumptions and assign someone to review performance and expiration dates.
Monitor utilization, coverage and net savings
Utilization measures the share of purchased commitment consumed at Savings Plans rates. In the quiet hour from the first example, $7 used out of $14 committed means 50% utilization. AWS utilization report.
Coverage measures the share of applicable usage costs receiving Savings Plans benefits. That same hour can have 100% coverage while losing money because half the commitment is unused. AWS coverage report.
Review both alongside net savings:
Net savings = comparable On-Demand cost − effective commitment cost − uncovered usage cost − applicable fees
High utilization with low coverage can justify investigating another purchase, but first confirm that the uncovered usage will persist. High utilization can also conceal weaker rate savings, as the account-ownership and migration examples demonstrate.
Our Savings Plans coverage vs utilization guide explains how to interpret the metrics before buying more.
Can you cancel an AWS Compute Savings Plan?
Treat a purchase as a commitment for its full term.
AWS offers a limited return process for eligible active plans with an hourly commitment of $100 or less, purchased within the previous seven days and returned within the same UTC calendar month.
Return quotas and other restrictions apply. This is a purchase-error remedy, not an ongoing cancellation option. AWS Savings Plan return conditions.
Evaluate your AWS commitments with Usage.ai
Usage.ai helps assess uncovered AWS usage and recommend commitments through its AWS cost optimization platform.
Start with a read-only Savings Test to evaluate the opportunity before granting purchasing permissions.
Eligible commitments managed under the Flex Insured Commitment Program include Cashback protection, subject to program terms. Fees are based on a share of realized savings from Flex Commitments.
Review eligibility, protection conditions and savings after fees when comparing the proposal with buying directly.
Frequently asked questions
What is an AWS Compute Savings Plan?
It is a one-year or three-year hourly spending commitment that provides discounted rates on eligible EC2, Fargate and Lambda usage. The commitment is measured in dollars per hour at Savings Plans rates.
Does one Savings Plan cover only one instance?
No. A Compute Savings Plan is an hourly dollar commitment, not a fixed number of instances or instance-hours. One plan can cover eligible usage across multiple instances and services until its commitment is consumed.
What happens if I add more EC2 instances after buying a plan?
New eligible usage enters the allocation process. The commitment remains fixed, and AWS decides which usage receives the benefit under its application rules.
Adding resources can shift coverage between workloads; it does not increase the purchased commitment. AWS automatic application rules.
Do Compute Savings Plans apply across AWS Regions?
Yes. Compute Savings Plans can apply across Regions and EC2 families, subject to eligibility. EC2 Instance Savings Plans remain tied to one selected family and Region. AWS plan flexibility.
Do AWS Compute Savings Plans cover RDS?
No. Compute Savings Plans cover eligible EC2, Fargate and Lambda usage. AWS offers separate Database Savings Plans for eligible database usage, including supported RDS usage. AWS Savings Plans types.
What happens to unused hourly commitment?
You pay for it, and it cannot roll forward to another hour. Choose the commitment using hourly demand and reduced-usage scenarios.
Does a Compute Savings Plan guarantee EC2 capacity?
No. Savings Plans provide a billing discount. Evaluate an On-Demand Capacity Reservation separately if you need capacity assurance. AWS capacity reservation guidance.