EC2 Idle & Rightsizing Calculator
See what idle and oversized EC2 instances cost, and compare stopping, downsizing or a Savings Plan.
Each one fills in the fields below for a typical workload of that kind. You can change any of them.
- Buying hours nobody needs
- $2,035.20/mo
- Storage, which a stop does not end
- $160.00/mo
- Best single fix
- $2,035.20/mo
The three fixes, priced
| Schedule it off outside needed hoursEBS volumes keep billing all month, running or stopped: $160.00/mo | $2,035.20/mo |
| Rightsize to m5.largePeak CPU of 18.0% leaves room for one size down, and to 8 GiB of memory from 16: check memory use first | $1,401.60/mo |
| Commit to a Savings PlanIndicative band, 1yr no-upfront to 3yr. Compute only, storage is never covered by a commitment. | $560.64 to $1,261.44/mo |
| Best single option | $2,035.20 |
And the carbon
Running this today: about 0.54 to 1.29 t CO₂e a year in us-east-1 (343 g/kWh).
- Schedule it off outside the needed hourssaves 392 to 939 kg CO₂e a year (73%)
- One size down, m5.largesaves 270 to 646 kg CO₂e a year (50%)
Reservations and savings plans changes the price, not the hardware or the hours, so it saves no carbon. An estimate of the servers’ electricity (compute and memory, with data center overhead), location-based, from Cloud Carbon Footprint’s coefficients; your provider’s own carbon tool is the authoritative figure. Estimate a whole workload or see the methodology.
What this model assumes
Show
- Storage is charged in every scenario. Only the compute line responds to scheduling. This is the difference between this tool and the ones that quote a percentage of your whole bill.
- Sizing down is suggested one step at a time. Halving vCPU roughly doubles how busy the instance is, so we suggest a smaller size only when peak CPU stays under 40.0%. Sizing down at 60% peak CPU risks an outage. Where a family has no size with fewer vCPU, the step is the same vCPU with less memory (t3.large to t3.medium): memory use decides that one, and this page does not see it.
- Savings Plan discounts are shown as a range. They vary by instance family, region, term and payment option, so one percentage would be wrong for most readers.
- Do not buy a Savings Plan for hours you plan to switch off. Scheduling and committing do not add up: pick the larger single saving, not the sum of both.
Prices verified 13 Oct 2026. Public list prices in USD for US East (N. Virginia), before any discount you have negotiated.
Want this done on your real bill?
Cloud Cost Audit: We go through your real AWS or Google Cloud bill and hand you a ranked list of what to cut first. From $1,500, fixed. Nothing paid upfront.
Questions
Do you get charged for a stopped EC2 instance?
Not for the instance hours, but yes for its storage. EBS volumes attached to a stopped instance bill at the full rate, so a stopped m5.xlarge with a 500 GiB volume still costs $40 a month. An Elastic IP keeps billing too: since February 2024 AWS charges $0.005 an hour for every public IPv4 address, attached to a running instance or not, so a stopped instance still pays about $3.65 a month for each Elastic IP it holds.
How much can I save by scheduling EC2 instances off?
A weekday 8am-8pm schedule runs about 260 hours instead of 730, so roughly 64% of the compute charge. That figure applies to compute only. Once storage is counted the real saving on a dev fleet with 100 GB disks is about 60%, and lower still for instances with large volumes.
Should I rightsize or buy a Savings Plan?
Rightsize first, then commit to what is left. A Savings Plan bought against capacity you later remove is wasted commitment you cannot cancel, and the same logic applies to scheduling, commitments cover hours, so committing to instances you plan to switch off is paying twice. This calculator deliberately shows the largest single option rather than a sum.
What CPU utilization counts as an idle EC2 instance?
Below about 5% peak CPU an instance is doing essentially nothing and should be questioned rather than optimized. Below 40% peak there is usually room for one size down, but not more than one, because halving vCPU roughly doubles utilization. CPU alone is not sufficient evidence: an instance can be memory-bound or network-bound at 5% CPU and still be busy.
Is it better to stop or terminate an idle EC2 instance?
Terminating releases the storage as well as the compute and is the only option that saves the whole cost. Stopping keeps the volumes, so it is the right choice for something you will genuinely restart and the wrong one for an instance nobody can name an owner for. If the peak CPU is near zero and nobody claims it, the answer is usually terminate.
Related tools
EBS gp2 vs gp3 Calculator
The easiest 20% saving in AWS, unless your volume needs paid IOPS.
OpenFargate vs EC2 Calculator
Fargate costs more per vCPU, but often less for a whole cluster.
OpenAmazon RDS Cost & Rightsizing Calculator
See the Multi-AZ premium in compute and in storage, then price a cheaper setup.
OpenMore tools for: Compute and utilization
Idle is easy to price and hard to find.
Finitizer profiles every instance against CPU, network and disk together, not CPU alone, so it distinguishes idle from memory-bound, and it tracks whether a commitment you already own still matches the fleet you actually run.
See how Finitizer works