1. Your usage
Use the on-demand price of what you ran, not what you were billed after existing discounts. Values are treated as equally spaced samples of a typical month. If you fill in the steady workload instead, the pasted usage is ignored.
2. The options you are choosing between
The example discounts are round numbers for illustration, not current prices. Replace them with the effective discount versus on-demand from your own AWS quote or Cost Explorer recommendation.
| Option | Term (months) | Discount vs on-demand (%) | Paid upfront (% of total) |
|---|---|---|---|
Leave it at 0 to compare nominal dollars. A positive rate discounts money paid upfront, which is what makes an all-upfront option less attractive than its headline discount.
Paste your usage, then click the button.
Result
| Option | Commit / hour | Covers | Used | Saves / month | Saves | Upfront payback | If usage −30% |
|---|
Savings against how much you commit (best option)
Stress test: what if usage changes after you commit? (best option)
| Usage vs what you pasted | Savings vs on-demand |
|---|
How it is calculated
Commitment model: a commitment of C dollars per hour is paid every hour. Usage priced at on-demand U costs U×(1−d) while covered, so C covers x = C/(1−d) of usage each hour and the rest is billed on demand: hourly cost = C + max(0, U − x). That cost is convex in x, and its minimum is where usage exceeds x exactly (1−d) of the time, so the optimum is the d-quantile of your usage. The tool evaluates every breakpoint in your data rather than assuming a shape. Cost of capital: a share f paid upfront over T months is converted to a present-value discount, 1 − (1−d)×(f×T/A + 1 − f), where A is the sum of monthly discount factors; at 0% it changes nothing. Payback: upfront paid divided by the monthly benefit of running the commitment. Stress: the chosen commitment is held fixed while every usage value is scaled.
Not modelled: Spot pricing, instance-size and region flexibility rules of a particular product, Reserved Instance marketplace resale or exchange, service-specific eligibility, taxes and credits, and prices changing during your term. Unverified: the maths is unit-tested against closed forms and brute-force search only; it has not been compared with any cloud provider's own recommendation engine, and AWS pricing is not fetched.
Common questions
Should I buy a Savings Plan or Reserved Instances?
Both trade a fixed hourly commitment for a discount, and the arithmetic is the same: you pay the commitment every hour whether you use it or not. Savings Plans are broader, applying across instance sizes, and in some cases families and services, while Reserved Instances are tied to an instance type and give up flexibility for a sometimes larger discount. Enter the effective discount from your own quote for each and this tool shows which saves more on your actual usage pattern.
How much should I commit to?
Less than your peak and usually less than your average. Each dollar of commitment is used only in the hours when usage is at least that high, and it pays off only if that happens more often than one minus the discount. With a 30% discount, cover the level your usage exceeds 70% of the time, which is the 30th percentile of hourly usage. The tool finds that level from your own data and shows the savings curve around it.
What is the break-even utilization of a commitment?
It equals one minus the discount. A commitment with a 30% discount breaks even if the committed capacity is used 70% of the time, and loses money below that. A 3-year commitment at a 50% discount breaks even at 50% utilization, but it locks that spend in for three years, so the risk is that usage falls, not the discount.
Is a 3-year commitment worth it?
It has the highest headline discount and the highest downside. The stress table shows your savings if usage were 50%, 70%, 85%, 115% or 130% of what you pasted. If savings turn negative at 70%, a 3-year term is a bet that your workload will not shrink. Committing to the always-on base and leaving the peaks on demand keeps most of the saving with much less regret.
Does paying upfront beat no upfront?
Only if the extra discount exceeds what the money could earn elsewhere. Enter your cost of capital and the tool converts each option to a present-value discount, so an all-upfront option with a slightly higher headline discount can lose to no-upfront when money is expensive. It also shows how many months of running the commitment it takes to recover the upfront payment.
Where do I get the hourly usage numbers?
Export hourly or daily on-demand-equivalent cost for the service from your cloud billing tool, one value per line, or a two-column CSV with the date first. Four or more weeks is better than one, and include a month-end if you have one. Daily totals work too; choose per-day in the dropdown.
Is my usage data uploaded anywhere?
No. Everything is computed in this browser tab. There are no accounts, cookies or analytics vendors, and no prices are looked up, so you type in the discounts from your own quote.