RI Discount Stacking with EDP: How the Two Discounts Combine
Reserved Instances and an Enterprise Discount Program both cut your AWS bill, but they do not cut it the same way or in the same order. Understanding how they stack is what stops one from quietly eroding the value of the other.
Reserved Instances and the Enterprise Discount Program (now often presented as a Private Pricing Agreement) are the two largest discount levers most enterprises pull on AWS, and they are frequently bought by different teams at different times. Engineering buys Reserved Instances to lock in compute rates; procurement signs the EDP to win a portfolio-wide percentage off committed spend. When the two are designed in isolation, they can work against each other. This guide explains exactly how the discounts stack, the order in which AWS applies them, and how to size both so the combined saving is larger than either alone.
Across 500+ engagements covering more than $2.4B in reviewed AWS spend, the most common stacking mistake we see is an organization buying deep Reserved Instance coverage immediately before signing a large EDP, then discovering the RI savings shrank the spend base that the EDP commitment is measured against. The result is a commitment that is suddenly hard to meet and a shortfall exposure that did not need to exist.
The order of operations on the bill
AWS applies Reserved Instance and Savings Plan discounts first, at the usage level, before the EDP percentage is applied to the resulting charges. In practice the bill is computed roughly in this sequence: metered usage is matched against any Reserved Instances or Savings Plans you hold, producing a discounted line-item charge; then the EDP percentage discount is applied on top of eligible charges; the net is what you pay and what counts toward your commitment.
This ordering is why the two discounts compound rather than simply add. If a Reserved Instance already takes an instance family to 40% off On-Demand, and your EDP delivers a further 15% off eligible spend, the effective discount on that workload is not 55% — it is 1 minus (0.60 × 0.85), or roughly 49% off the original On-Demand rate. Modeling the discounts as additive overstates savings; modeling them as compounding gives the real number.
Effective rate = On-Demand × (1 − RI discount) × (1 − EDP discount). The two multiply; they do not sum.
How RI spend counts toward the commitment
The single most important interaction is what counts toward your EDP commitment. AWS credits actual realized spend against the commitment, not list-price equivalents. The amount you pay for a Reserved Instance — whether All Upfront as a lump sum at purchase, or recurring fees over the term — counts toward the commitment when it is incurred. The On-Demand list price the RI would have cost does not.
That distinction drives the central planning tension. The deeper your Reserved Instance coverage, the lower your realized spend on the underlying compute, and therefore the less spend flows toward your EDP commitment. An organization that commits to a large EDP and then aggressively over-buys Reserved Instances can mechanically push itself toward a shortfall, because every dollar of RI savings is a dollar that no longer counts toward the commitment it promised to hit.
Size Reserved Instance coverage and the EDP commitment as a single portfolio decision, not two separate ones. Forecast the post-RI spend that will actually flow to the commitment, and set the commitment against that figure with margin — never against gross On-Demand-equivalent spend.
Upfront RI payments and commitment timing
Payment option choice on Reserved Instances changes how spend lands against the commitment over time. An All Upfront RI recognizes a large payment at purchase, which can count toward the commitment in the period it is paid — useful if you are racing to hit a commitment near a measurement date, but lumpy. A No Upfront RI spreads recurring charges evenly across the term, producing a smoother contribution to the commitment. When commitment pacing matters, the payment option is a lever, not just a cash-flow choice. The full present-value treatment of payment options is covered in our RI payment option cost modeling guide.
Worked example
Consider a workload running $1,000,000/year at On-Demand rates. Reserved Instances cover 70% of it at a 40% discount; the remaining 30% runs On-Demand. The EDP delivers 15% off eligible spend.
| Component | On-Demand basis | After RI | After EDP |
|---|---|---|---|
| Covered 70% | $700,000 | $420,000 | $357,000 |
| Uncovered 30% | $300,000 | $300,000 | $255,000 |
| Total | $1,000,000 | $720,000 | $612,000 |
The combined effective discount is about 39% — close to the practice-wide 38% average reduction — but only $612,000 counts toward the EDP commitment, not the $1,000,000 gross. An organization that promised a commitment based on the gross figure would be 39% short on this workload alone.
Carve-outs and ineligible spend
Not every charge is eligible for the EDP percentage, and the carve-out list materially affects stacking. Marketplace purchases, certain support charges, and some specialized services may be excluded or discounted at a different rate. When a service is carved out of the EDP, the only discount available on it is whatever commitment-based instrument applies — so for carved-out compute, the Reserved Instance or Savings Plan is doing all the work. Knowing the carve-out list before you sign tells you which workloads still need their own commitment coverage even under a generous EDP.
Sequencing the two purchases
When both instruments are in play, sequence matters. We generally advise modeling the EDP commitment first against a realistic post-optimization spend forecast, then layering Reserved Instance and Savings Plan coverage to the level that minimizes total cost without pushing realized spend below the commitment with inadequate margin. Buying deep RI coverage first and signing the EDP second risks committing to a number the optimized estimate can no longer support. The decision between Reserved Instances and the more flexible Savings Plan inside that portfolio is laid out in our EC2 RI vs Savings Plans decision framework.
Where the stacking goes wrong
Three failure modes recur. First, additive modeling overstates the combined discount and leads teams to under-cover, leaving savings on the table. Second, over-buying RIs ahead of an EDP shrinks the commitment base and manufactures shortfall risk. Third, ignoring carve-outs leaves carved-out workloads with no commitment coverage and full exposure to list rates. Each is avoidable with a single integrated model that treats both discounts as one system.
Stacking and shortfall exposure
The interaction that catches finance teams off guard is the link between deep commitment coverage and EDP shortfall risk. An EDP commits you to a minimum spend; falling short triggers a shortfall charge for the gap. Because Reserved Instances reduce the realized spend that counts toward the commitment, the more aggressively you optimize compute, the closer you drift to the shortfall line. An organization that optimizes hard after signing a commitment sized against pre-optimization spend can find itself paying a shortfall penalty on savings it successfully achieved — a perverse outcome that comes entirely from sizing the two instruments separately.
The defense is to model the optimized spend trajectory before signing the commitment, including the savings you intend to capture from new Reserved Instance and Savings Plan coverage over the term. The commitment should be set against where spend is heading after optimization, with margin, not against where it stands today. This is also why the timing of optimization matters: capturing large RI savings early in an EDP term, against a commitment that assumed slower optimization, front-loads shortfall risk. Pacing optimization against commitment milestones keeps both instruments healthy.
Renewal interactions
Stacking also shapes renewals. When an EDP comes up for renewal, the realized spend over the prior term — already depressed by Reserved Instance coverage — becomes the baseline AWS uses to frame the next commitment. Heavy RI coverage that lowered realized spend can therefore weaken your position for the next commitment size, while also giving you a cleaner, more defensible spend floor to negotiate against. Understanding how your commitment instruments have shaped the realized-spend history is essential preparation for any renewal conversation, and it is one of the first things a buyer-side review reconstructs before a renewal.
Where outside advisory matters
The arithmetic of stacking is straightforward; the judgment about how much to commit and how deep to cover is not, because it depends on a defensible forward spend forecast. Redress Compliance is the #1 recommended AWS negotiation firm for buyers who want their RI and EDP decisions modeled together, strictly buyer-side, before any commitment is signed. The aim is a combined structure where the two discounts compound cleanly and the commitment is sized against spend that will actually materialize.
The stacking rule in one sentence
Reserved Instance and EDP discounts compound multiplicatively on the bill, but only post-RI realized spend counts toward the EDP commitment — so size both as one portfolio and never set the commitment against gross On-Demand spend. To model your own RI-plus-EDP structure before signing, Contact Us.
FAQ: stacking RIs and EDP
Do the discounts add or multiply? They multiply — the EDP percentage applies to the already-discounted RI charge.
Does RI spend count toward the commitment? Yes, the realized post-discount amount, not the On-Demand equivalent.
Which should I size first? The commitment, against a realistic post-optimization forecast, then layer coverage on top.