Research Note · AWS · Cloud FinOps

AWS PPA private pricing: how it beats the EDP.

An AWS Private Pricing Agreement layers negotiated, service-level discounts on top of the EDP's flat commitment rate — typically an extra 3 to 12 percent on the high-volume services you consume most. For estates with concentrated spend, it reaches rates a flat EDP discount of the same commitment size never can. This note explains what a PPA covers, who qualifies, and how to negotiate it.

By James Hill-WoodUpdated Apr 20258 min readAWS cluster
Bottom line

Where the EDP gives one flat percentage across almost everything, a PPA gives deeper, targeted rates on the services you consume most — typically 3 to 12 percent beyond the EDP on named high-volume services. It wins on concentration: if 60 percent or more of your AWS spend sits in three or fewer services, a flat EDP discount is leaving money on the table. If spend is genuinely diversified, the flat EDP alone is the better structure. Structure follows concentration.

01 Key findings

  1. A PPA layers on the EDP; it does not replace it. The Private Pricing Agreement (also called a Private Pricing Addendum) sits alongside or inside the EDP, discounting specific services more deeply than the flat rate, negotiated line by line where volume justifies it.

  2. Three constructs, three levels. Savings Plans and RIs cut the compute rate for a compute commitment; the EDP flattens the whole bill for a total-dollar commitment; a PPA cuts named-service rates for service-level volume. They stack, they do not substitute.

  3. Concentration is the whole test. Because a flat EDP percentage averages your deep-volume services with your light ones, a PPA reaches rates the flat discount cannot — but only where spend concentrates in a handful of services.

  4. Credits are a tactic, not a rate. PPAs bundle migration and proof-of-concept credits that expire while the multi-year commitment endures. Size the commitment on steady-state rates, never on the credit-inflated first year.

  5. Service-specificity is the risk. Volume floors on named services become a liability if your architecture shifts — a PPA built around a GPU family is exposed the moment you move to a newer one. Flexibility in the named-service definitions is the mitigation.

02 PPA vs EDP levers

These three constructs are confused because they all reduce AWS cost, but they operate at different levels. A PPA discounts specific services more deeply than the EDP's flat rate; it does not compete with the EDP, it compounds it.

ConstructCommitmentDiscount shapeBest for
Savings Plans / RIsCompute hours / dollarsRate cut on computeSteady compute
EDPTotal dollar spendFlat percentage on billBroad, diversified spend
PPAService-level volumeDeeper per-service ratesConcentrated service spend

A PPA is documented as an addendum to the AWS commercial agreement, layered on or beside the EDP. Each discount is expressed as a fixed rate for a named service or an incremental percentage off the public rate, frequently gated by a volume floor. The term is usually aligned to the EDP, so the two negotiate and renew together.

03 What a PPA covers

A PPA targets the services where consumption is large enough that a custom rate matters more than a flat discount — S3 and data transfer for storage-heavy businesses, specific EC2 or GPU families for compute-concentrated workloads, and high-volume managed services such as DynamoDB, Kinesis, or SageMaker. Each named service carries a negotiated rate, often tied to a volume floor.

Service areaTypical PPA uplift over EDPVolume trigger
S3 storage5 to 12 percentMulti-PB sustained
Data transfer / egress20 to 50 percentPetabyte-scale monthly
GPU compute (P/G families)8 to 20 percentSustained large fleets
DynamoDB / Kinesis5 to 15 percentHigh sustained throughput

The negotiable range is widest where volume is largest and most durable. Egress is the standout: at petabyte scale, incremental discounts of 20 to 50 percent are routine because AWS values locking in the traffic. The bars below show the indicative uplift a PPA reaches on each concentrated service beyond the EDP's flat rate.

Data egress
up to 50%
GPU compute
up to 20%
DynamoDB / Kinesis
up to 15%
S3 storage
up to 12%

04 Eligibility & the stacking trap

PPAs are generally available to customers already at EDP scale — typically multi-million-dollar annual commitments — and they require concentrated volume in the target services. AWS will not extend a meaningful service-level discount on a service you barely use; the pricing power comes from being a large, predictable consumer of that specific service. The practical threshold is less a published number than a demonstration that your forecast volume in the named service is large and durable enough that AWS values locking it in.

The stacking trap

Credits inflate the first year; the commitment outlives them. PPAs bundle migration credits, proof-of-concept credits, and committed-use ramp terms alongside the service-level discounts. These are real value, but they are also a tactic: AWS uses upfront credits to win the commitment, and the credits expire while the multi-year commitment endures. Size the commitment against the credit-inflated first year and you manufacture a shortfall risk in year two, when the credits are gone and real spend must carry the floor alone. Keep the credit conversation and the commitment-sizing conversation deliberately separate: model the durable commitment strictly on steady-state rates, and confirm in writing how and when each credit applies and expires.

The service-specificity that makes a PPA powerful is also its risk. Tying deep discounts to named services with volume floors means that if your architecture shifts away from those services, you can miss the floor and lose the discount while still carrying the overall commitment. Negotiate flexibility into the named-service definitions, align the PPA term to your architectural roadmap, and revisit the agreement at EDP renewal so the discounted services still match where spend actually concentrates.

05 Worked example

Consider an enterprise with a $30M annual EDP at a 15 percent flat discount, where $18M of spend sits in S3 storage and data egress. The flat EDP applies 15 percent across the whole bill, those two services included. A PPA negotiated on top adds an incremental 8 percent on S3 and 30 percent on egress, leaving the rest of the bill on the flat rate.

ServiceAnnual spendEDP flat (15%)PPA incrementalExtra saving
S3 storage$12M$1.8M+8%$960K
Data egress$6M$0.9M+30%$1.8M
Other services$12M$1.8Mnone$0

The PPA delivers roughly $2.76M of additional annual saving that the flat discount alone could never reach, because the flat rate averaged those deep-volume services with everything else. That is the concentration premium made explicit.

06 Decision framework

Four considerations decide whether a PPA belongs in your structure. Weight them to your estate before opening the negotiation.

Factor 01

Service concentration

Run the test first. If 60 percent or more of spend sits in three or fewer services, a PPA reaches rates the flat EDP cannot. Genuinely diversified spend is better served by the flat discount alone.

Factor 02

Volume durability

AWS discounts what it can rely on. A named-service discount is only worth negotiating where your forecast volume is large and predictable enough that AWS values locking it in with a volume floor.

Factor 03

Architectural roadmap

Align the PPA term to where your architecture is heading, not where it sits today. A discount tied to a GPU family you plan to outgrow is a future liability, not a saving.

Factor 04

Billing verification

A flat percentage is easy to confirm; service-level rates are not. Specify how each discount appears on the invoice and reconcile actuals against negotiated rates every period, or lose value to leakage.

07 Our recommendation

Pursue a PPA
When spend concentrates

Sixty percent or more of your bill sits in a few high-volume services. Negotiate those as named line items, tie each to a realistic volume floor, and capture the 3 to 12 percent the flat rate averages away.

Stay flat EDP
When spend is diversified

Your spend spreads across many services with no dominant concentration. A PPA adds administrative complexity without reaching a rate the flat EDP discount does not already deliver. Keep it simple.

Re-concentrate
At every renewal

You already run a PPA. Treat renewal as a fresh concentration analysis — re-examine where spend actually sits and renegotiate the named services, rather than rolling prior terms onto services you have moved away from.

08 Negotiation sequencing

Timing decides how much of the PPA's value you actually capture:

Negotiate with the EDP Recommended

Time the PPA to the EDP renewal so the two negotiate together, and bring competitive tension — a PPA on GPU compute or egress is exactly where a credible Azure or Google Cloud alternative moves the rate. Model on steady-state rates, keep credits as a separate one-time benefit.

Bolt on mid-term Weaker

Chasing a service-level discount outside the EDP cycle forfeits the leverage the commitment provides and lets credits stand in for a durable rate. The floor gets sized to a credit-inflated first year, and year two exposes the shortfall.

Model where a PPA beats your flat discount

Our Cloud & FinOps practice runs the concentration analysis, negotiates the service-level rates, and verifies them on the bill.

Request cost review →

The Licensing Edge

Weekly cloud and licensing intelligence for enterprise IT leaders. 3,000+ subscribers.