Savings Plans vs Rightsizing: Buy the discount after you cut waste
Rightsize before you buy Savings Plans: AWS bills a Savings Plan’s hourly commitment for every hour of its 1- or 3-year term whether you use it or not, so a plan bought before cleanup keeps charging you for the capacity you delete. The right order is cut, then commit: delete idle resources, rightsize running instances, schedule non-production environments, and only then buy a Savings Plan sized to the stabilized usage.
Why should I wait to buy Savings Plans?
A Savings Plan commits you to a fixed hourly charge for 1 or 3 years, regardless of actual usage. AWS applies that commitment to eligible On-Demand usage, EC2, Fargate, Lambda, each hour. If your eligible usage drops after purchase, you still pay the full commitment. Rightsizing after you commit doesn’t reduce that obligation.
Cost Explorer reports the gap as unused commitment. In the Cost and Usage Report, the gap is the difference between savings_plan_total_commitment_to_date and savings_plan_used_commitment on SavingsPlanRecurringFee rows.
At a former employer, I cut a roughly $1M-a-year cloud bill by 26%. Whatever you cut, a Savings Plan bought before you cut it keeps billing you for the removed usage until the term ends.
How do I find my current Savings Plan utilization?
In the AWS Billing and Cost Management console, go to Savings and Commitments > Savings Plans. The Utilization report shows the percentage of your hourly commitment that was actually used. Coverage shows what share of your eligible On-Demand spend was covered by the plan.
Run this AWS CLI command to see daily utilization and unused commitment:
aws ce get-savings-plans-utilization \
--time-period Start=2026-10-01,End=2026-11-01 \
--granularity DAILY \
--query 'SavingsPlansUtilizationsByTime[].[TimePeriod.Start,Utilization.UtilizationPercentage,Utilization.UnusedCommitment]' \
--output table
Check coverage to confirm your plan is still covering the right mix:
aws ce get-savings-plans-coverage \
--time-period Start=2026-08-01,End=2026-11-01 \
--granularity MONTHLY \
--query 'SavingsPlansCoverages[].[TimePeriod.Start,Coverage.CoveragePercentage,Coverage.OnDemandCost]' \
--output table
Check your own bill in 2 minutes: https://revelations.consulting/checker?utm_source=blog&utm_medium=organic&utm_content=blog-savings-plans-vs-rightsizing
How do I cut waste first and then buy Savings Plans safely?
Follow this sequence:
- Delete idle resources. Cost Explorer shows cost, not activity, so use CloudWatch and AWS Compute Optimizer to confirm an instance is idle. Thirty days of daily maximum CPUUtilization near zero makes an instance a candidate. If the instance belongs to an Auto Scaling group, lower the group’s capacity instead, or the group will replace it.
Stopping an instance erases its instance store volumes, and terminating it also deletes every EBS volume with DeleteOnTermination=true. Copy any instance store data off while the instance is still running, then stop it and snapshot its EBS volumes. Terminate only after aws ec2 describe-snapshots shows every snapshot in the completed state.
# Check daily max CPU over 30 days
aws cloudwatch get-metric-statistics \
--namespace AWS/EC2 --metric-name CPUUtilization \
--dimensions Name=InstanceId,Value=i-0123456789abcdef0 \
--start-time 2026-10-13T00:00:00Z --end-time 2026-11-12T00:00:00Z \
--period 86400 --statistics Maximum \
--query 'sort_by(Datapoints,&Timestamp)[].[Timestamp,Maximum]' --output table
# Snapshot all EBS volumes on the instance
aws ec2 create-snapshots \
--instance-specification InstanceId=i-0123456789abcdef0 \
--description 'pre-termination backup'
- Rightsize running instances. Compute Optimizer bases EC2 recommendations on 14 days of CloudWatch metrics by default. Memory is included only if the CloudWatch agent publishes it. Without the agent, a recommended downsize can run the instance out of memory, so check memory yourself before you act.
Changing instance types requires a stop and start, so plan downtime, and expect the public IP to change unless you use an Elastic IP. Confirm the new type supports your AMI’s architecture (Graviton needs arm64). When moving to a Nitro family, also confirm the AMI has ENA and NVMe drivers.
aws compute-optimizer get-ec2-instance-recommendations \
--filters name=Finding,values=Overprovisioned \
--query 'instanceRecommendations[].[instanceArn,currentInstanceType,recommendationOptions[0].instanceType]' \
--output table
- Schedule non-production environments. Use Instance Scheduler or EventBridge Scheduler to stop dev and test instances outside business hours. For Auto Scaling groups, set desired and minimum capacity to zero instead of stopping instances. Tag environments (e.g., Environment=dev) and use Cost Explorer to quantify the opportunity.
- Commit only to the steady floor. Wait until your usage stabilizes after cleanup. Your floor is the lowest hourly eligible spend that holds every hour, including nights and weekends. You want a commitment that floor uses in full.
A Savings Plan commitment is stated in discounted $/hour, not On-Demand $/hour, so do not enter your On-Demand floor as the commitment. Let the purchase recommendation do that conversion, then check its estimated utilization.
The recommendation sizes the commitment for maximum estimated savings, not for your lowest hour, so it can exceed your floor. If EstimatedAverageUtilization is below 100, the recommended commitment goes unused in some hours. In that case, buy less than HourlyCommitmentToPurchase:
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 THIRTY_DAYS \
--query 'SavingsPlansPurchaseRecommendation.SavingsPlansPurchaseRecommendationDetails[].[HourlyCommitmentToPurchase,EstimatedAverageUtilization,EstimatedMonthlySavingsAmount]' \
--output table
Run this only after 30 full days have passed since your last cleanup change, or switch to --lookback-period-in-days SEVEN_DAYS if a week of post-cleanup usage is representative. A lookback that includes pre-cleanup data will overstate your need.
Compute Savings Plans cover EC2 across families, Regions, OS, and tenancy, plus Fargate and Lambda. EC2 Instance Savings Plans offer a deeper discount but are locked to one instance family in one Region. Compare rates on https://aws.amazon.com/savingsplans/compute-pricing/ (checked November 2026).
What it usually looks like
The first pattern is a Compute Savings Plan bought before cleanup. After rightsizing, eligible usage drops but the hourly commitment stays the same. Savings Plans utilization drops in the month of the cleanup, and Cost Explorer reports unused commitment for every hour that eligible usage stays below the commitment, until the term ends.
You can see this in your own Cost and Usage Report. Look for the SavingsPlanRecurringFee line item. The gap between total commitment and used commitment grows after a rightsizing event. Replace cur_table with your CUR table name:
SELECT date_trunc('month', line_item_usage_start_date) AS month,
savings_plan_savings_plan_a_r_n AS savings_plan,
SUM(savings_plan_total_commitment_to_date) AS committed,
SUM(savings_plan_used_commitment) AS used,
SUM(savings_plan_total_commitment_to_date) - SUM(savings_plan_used_commitment) AS unused
FROM cur_table
WHERE line_item_line_item_type = 'SavingsPlanRecurringFee'
GROUP BY 1, 2
ORDER BY 1, 2;
The second pattern is an EC2 Instance Savings Plan stranded by a migration. An EC2 Instance Savings Plan covers one instance family in one Region, such as m5 in us-east-1. If you move those workloads to Graviton instances or to Fargate, the commitment has nothing left to apply to. List your active EC2 Instance Savings Plans with their family, Region and end date:
aws savingsplans describe-savings-plans \
--states active \
--query 'savingsPlans[?savingsPlanType==`EC2Instance`].[savingsPlanId,ec2InstanceFamily,region,commitment,end]' \
--output table
Then check utilization per plan:
aws ce get-savings-plans-utilization-details \
--time-period Start=2026-10-01,End=2026-11-01 \
--query 'SavingsPlansUtilizationDetails[].[SavingsPlanArn,Utilization.UtilizationPercentage,Utilization.UnusedCommitment]' \
--output table
Repeat the command with each month's dates, for example Start=2026-09-01,End=2026-10-01 and Start=2026-08-01,End=2026-09-01, and compare the results per plan. If a plan's utilization dropped in the month you moved off its family, the commitment is stranded until its end date. Finish family and Fargate migrations before you buy an EC2 Instance Savings Plan. If a migration is still planned, use a Compute Savings Plan, which follows the workload.
If you want a second pair of eyes on your utilization, coverage and cleanup order before you sign a commitment, book a free 30-minute consult with a senior engineer: https://revelations.consulting/?utm_source=blog&utm_medium=organic&utm_content=blog-savings-plans-vs-rightsizing#contact
Run your own numbers: Savings Plan Sizer.
Frequently asked questions
Should I buy Savings Plans or Reserved Instances?
For EC2, Fargate and Lambda, Compute Savings Plans are the most flexible commitment: they apply across instance families, sizes, Regions, OS and tenancy, while regional Linux Reserved Instances flex only by size within one family in one Region. Standard Reserved Instances and EC2 Instance Savings Plans usually discount more than Compute Savings Plans, so they fit only usage you are sure will stay on one instance family. Compute Savings Plans do not cover RDS or ElastiCache, so size database commitments separately, after rightsizing the databases, using the commitment options on each service’s pricing page.
Can I cancel a Savings Plan?
No. A Savings Plan cannot be cancelled, reduced, or resold, and AWS allows a return only within 7 days of purchase, in the same calendar month, and only for a small hourly commitment (the limits are in the AWS Savings Plans User Guide under returning a purchased Savings Plan). After that window, the full hourly commitment of a Savings Plan is billed for the whole term regardless of usage.
How often should I review my Savings Plan utilization?
Review Savings Plans utilization and coverage at least monthly, and after any major change such as rightsizing, a migration, or a scaling event. If unused commitment appears for two consecutive months, stop buying new plans. You cannot reduce an existing commitment.
Do Savings Plans apply to Spot Instances?
No. Savings Plans discount only eligible On-Demand usage, so Spot Instances are billed at the Spot price and get no Savings Plan discount. Use Spot for interruptible work on top of your Savings Plan floor.
Is it worth buying a 3-year Savings Plan?
A 3-year Savings Plan is worth buying only for the part of your usage you are confident will hold for three years, because the deeper rate comes with three times the term of a 1-year plan and, at the same hourly commitment, three times the total obligation. Start with 1-year Savings Plans on your cleaned-up floor, and add 3-year tranches only after that floor has held for several months.
Run the numbers on your own bill: How much should you commit to a Savings Plan?. Free, in your browser.
Want this kind of thing weekly? Get Cloud Bill Weekly. Want us to find it in your bill? We're taking three founding clients for a free Cost Assessment.