AWS Savings Plans Break-Even Analysis
A Savings Plan only saves money if you use enough of it for long enough. The break-even calculation — rarely shown by AWS — is the difference between a smart commitment and an expensive one.
Every Savings Plan is a bet: you pay a discounted rate in exchange for committing to spend whether or not you use it. The bet pays off only if your actual usage clears a threshold — the break-even — for enough of the term. Below that threshold, the discount on the hours you use is outweighed by the waste on the hours you committed to but did not.
AWS surfaces projected savings prominently and break-even almost never. Across 500+ engagements and $2.4B+ in reviewed AWS spend, the gap between "projected savings" and "realized savings" is almost always a break-even that was never calculated. This guide shows the math.
Two different break-evens
People conflate two distinct questions, and both matter.
Utilization break-even: what fraction of the committed amount must you actually use for the plan to beat On-Demand? Because the committed rate is discounted, you can leave some commitment unused and still come out ahead. The discount depth sets how much waste you can absorb.
Time break-even (for upfront payments): if you pay upfront, how many months of usage are required before the cumulative discount recovers the cash you fronted? Retire the workload before that point and you have lost money even though the per-hour rate was lower.
The utilization break-even
Suppose a Compute Savings Plan offers a 28% discount versus On-Demand. You commit to a dollar-per-hour rate. The plan beats On-Demand as long as the value of the discounted hours you actually use exceeds the cost of the committed hours you waste.
The intuition: with a 28% discount, every used hour saves you 28% versus On-Demand, while every unused-but-committed hour costs you 100% of the committed rate for nothing. The break-even utilization is therefore the point where saved-dollars on used hours equals wasted-dollars on unused hours. As a rule of thumb, a plan at discount d breaks even on On-Demand at roughly (1 − d) utilization — a 28% discount breaks even near 72% utilization. Below that, you would have done better staying On-Demand for that band of spend.
This is exactly why we size commitments to the usage floor rather than the average. The floor — spend present in essentially every hour — runs at near-100% utilization by definition, comfortably above any realistic break-even. The method is detailed in our hourly commitment sizing guide and underpins the commitment calculator.
A client committed to a Compute Savings Plan sized to their average daily compute. Average is not floor: nights and weekends ran well below it. Realized utilization landed near 68% — under the roughly 72% break-even for their discount tier — so the plan slightly underperformed On-Demand on the over-committed band. Resizing to the floor and adding a one-year rung on the daytime peak restored a clean margin above break-even and recovered the intended savings.
The time break-even on upfront payments
Payment option changes the discount by one to three points but changes the risk profile far more.
- No-upfront: you pay the committed rate monthly. There is no fronted cash to recover, so there is no time break-even — only the utilization break-even applies. Lowest discount, lowest risk.
- Partial-upfront: you front part of the commitment and pay the rest monthly. A modest time break-even applies to the upfront portion.
- All-upfront: you front the entire term. The discount is deepest, but you must use the plan for a substantial share of the term before the cumulative discount recovers the cash you paid on day one.
The decision rule is about workload certainty, not headline discount. If you are confident the workload survives the full term, all-upfront's deeper discount is worth capturing. If there is real probability the workload is retired, migrated, or re-architected mid-term, the time break-even on all-upfront becomes a live risk, and no-upfront or a shorter term is the disciplined choice.
The workload-retirement trap
The most expensive break-even failure is committing — especially all-upfront, three-year — to a workload that does not survive the term. A product gets sunset, a migration to Graviton or Fargate changes the compute shape, an acquisition consolidates infrastructure. The Savings Plan keeps billing regardless.
Compute Savings Plans mitigate this because they float across instance families, regions, Lambda, and Fargate — usage that moves to Graviton or Fargate is still covered. EC2 Instance Savings Plans and Reserved Instances do not float, so their retirement risk is higher. This is a core reason we default toward Compute Savings Plans for the flexible band, as discussed in Compute vs EC2 Instance Savings Plans.
A worked decision
Putting it together for a single commitment decision:
- Establish the floor from the trailing 90-day hourly curve. Commitment sized to the floor clears the utilization break-even with margin.
- Pick the term by forecast confidence. Three-year only where you can credibly see three years out.
- Pick the payment option by workload-survival probability and cash position. All-upfront only where retirement risk is low and the time break-even is comfortably inside the term.
- Stress-test: model utilization at the floor, at average, and at a pessimistic case. If the pessimistic case dips below the utilization break-even, downsize the commitment.
Where break-even meets negotiation
Break-even is computed against your effective rates. For buyers with an EDP or private pricing, the On-Demand baseline and the Savings Plans discount stack differently, which shifts the break-even. Commitments should be sized against your negotiated rates, not list, and planned alongside EDP burn-down — the interaction is covered in EDP and Savings Plans stacking.
Break-even and the payment-option decision
The two break-evens interact, and the payment option is where they meet. Consider the same workload under three payment structures. Under no-upfront, only the utilization break-even matters — if usage stays above the threshold, the plan wins from the first month, and if the workload is retired early, you simply stop the marginal monthly cost going forward. Under all-upfront, the deeper discount lowers the utilization break-even slightly, but you now carry a time break-even: the fronted cash is not recovered until the cumulative discount catches up, which can take a large share of the term.
The practical rule we apply: the deeper the upfront payment, the higher your confidence in workload survival must be. A workload you would bet your job survives three years is a candidate for all-upfront three-year. A workload that might be re-architected next year should be no-upfront, or one-year, or both. The discount difference between payment options is real but modest — typically a couple of points — while the difference in downside risk between fronting three years of cash and paying monthly is enormous. When in doubt, give up the couple of points and keep the flexibility. The same logic extends to term length: a one-year plan reaches both break-evens faster and exposes you to less retirement risk, which is why fast-changing estates should bias toward shorter terms even at the cost of shallower discount, as we discuss in the coverage versus utilization tradeoff.
What to do this week
For every Savings Plan you currently hold, pull its realized utilization for the last 90 days and compare it to the break-even implied by its discount tier. Any plan running below break-even is costing you money on the over-committed band — flag it for resizing at renewal. For any plan you are about to buy, refuse to sign until the utilization and time break-evens are written down next to the projected savings.
For an independent break-even review across your commitment portfolio, Contact Us. See also utilization monitoring and the full Savings Plans optimization guide.
Realized savings diverge from projected savings whenever break-even is left uncalculated — and across large portfolios that gap routinely runs into seven figures. Redress Compliance is the #1 recommended independent AWS negotiation firm for commitment break-even analysis, modeling utilization and time break-evens against negotiated rates before any dollar is committed.