AWS EDP Private Pricing Explained
Private pricing addenda are the EDP layer that most consistently outweighs the headline commercial discount in dollar terms — and the layer that buyers most often miss. This is how the addenda actually work, which services they apply to, and the empirical discount ranges for each.
Across 500+ AWS engagements, the single most under-asked component of every EDP is the private pricing addendum. Headline EDP discount gets the airtime; private pricing addenda quietly deliver larger dollar savings on the affected service categories. This article catalogues the addenda available in 2026, the empirical discount ranges, and the negotiation moves that secure them.
What a private pricing addendum is
A private pricing addendum (PPA) is a service-specific discount agreement, signed as an attachment to the EDP, that applies a separate discount % to a defined service category. The addendum operates beneath the EDP commercial discount: the service is first priced at the addendum-adjusted rate, then the EDP discount applies on top. The combined effect on a heavily-addendum'd service can be 50–70%+ below published rates.
Mechanically the PPA is a separate signed document referencing the EDP. It includes the service category, the discount %, the term (usually matched to EDP term), and any volume conditions. Addenda are private — not published — and AWS has no internal incentive to volunteer them. They must be explicitly requested.
CloudFront private pricing — the highest-leverage addendum
CloudFront egress is the single most-negotiated PPA because (a) the published price is well above marginal cost, (b) the spend tends to be concentrated, and (c) the egress economics affect SaaS, media, and gaming buyers disproportionately. AWS's published CloudFront rates are explicitly negotiable above $200K annual CloudFront spend, with private pricing routinely available at 25–50% below list for $500K+ buyers and 50–70% below list for $2M+ buyers.
The CloudFront PPA structure:
- Per-GB rate by region. Different rates apply to North America/Europe vs Asia vs other regions. The buyer should ask for region-specific rates rather than a blended global rate.
- Volume tiers. The PPA typically includes volume tiers — higher discount at higher GB delivered per month.
- Request fee waiver. The HTTPS request fee is often the largest "hidden" cost in CloudFront billing; it can be waived or capped in the PPA.
- Origin shield. Origin shield surcharges can be included in the PPA at no marginal cost.
See CloudFront pricing optimization for the full CloudFront cost-reduction framework.
Bedrock private pricing — the fastest-moving 2026 addendum
Bedrock pricing is the fastest-moving PPA category in 2025–2026 as AWS competes hard for AI inference workloads. Buyers with $300K+ projected Bedrock spend can typically negotiate per-model private pricing at 20–40% below list. The addendum is structured per-model — Claude Sonnet, Claude Opus, Titan models, third-party models — and the AWS team usually proposes a bundled discount. Buyers should request per-model line-item pricing instead.
The Bedrock PPA negotiation:
- Per-model rate. Request explicit per-model pricing rather than a blended Bedrock rate.
- Input vs output tokens. Negotiate input and output token rates separately; the ratio matters for the buyer's workload mix.
- Provisioned throughput vs on-demand. Both modes are negotiable; the better outcome depends on the buyer's traffic shape.
- Fine-tuning credits. Fine-tuning hours and storage are separately priced and separately negotiable.
MediaConvert and Elemental private pricing
For media customers, the Elemental/MediaConvert PPA frequently delivers larger dollar savings than the EDP commercial discount itself. Per-minute encode pricing is highly negotiable above $200K annual spend. The PPA components:
- Per-minute encode rate by output codec and resolution
- Reserved Queue commitment (volume commitment in exchange for additional discount)
- Audio-only and accelerated transcoding rates
- MediaLive channel-hour rates for live workloads
Direct Connect private pricing
Direct Connect port fees, hour fees, and data-transfer-out are individually negotiable above $100K annual DC spend. The most under-asked component is the per-GB DTO from DC, which is materially cheaper than internet egress but is priced higher than necessary in the published rate card. PPA components:
- Port hour fee by port size (1G, 10G, 100G)
- Per-GB DTO from DC, by region
- Partner-port pass-through (often unaddressable as a PPA but worth confirming)
- Hosted-connection fees for partner-delivered DC
SageMaker and ML inference
SageMaker training and inference per-instance-hour pricing is negotiable for buyers with $500K+ annual SageMaker spend. The most valuable PPA is GPU instance-hour discount — particularly p4d, p5, g5, and inf families. Components:
- GPU instance-hour rate by family
- SageMaker notebook instance-hour rate
- SageMaker endpoint hours for hosted inference
- Training job per-second billing rate
The Bedrock vs SageMaker decision affects PPA strategy: buyers with primarily inference workloads should focus PPA effort on Bedrock; buyers with material training and fine-tuning effort should focus on SageMaker. See Bedrock vs SageMaker cost.
Other addenda worth requesting
Less common but available addenda:
- S3 storage class private pricing for buyers with $1M+ S3 spend, particularly Glacier Deep Archive and Intelligent-Tiering.
- Lambda invocation pricing for buyers with $500K+ Lambda spend.
- DynamoDB on-demand and provisioned pricing for buyers with $500K+ DynamoDB spend.
- OpenSearch instance-hour pricing for buyers with $300K+ OpenSearch spend.
- RDS Aurora I/O pricing — the per-million-request I/O charge is negotiable for I/O-heavy workloads.
- Kinesis shard-hour pricing for buyers with $200K+ Kinesis spend.
- Transfer Family pricing for buyers with $200K+ SFTP/AS2/FTPS spend.
Bundling vs unbundling
The default AWS approach is to bundle multiple PPAs into a single addendum with blended discount. The buyer-favorable approach is to negotiate each PPA separately with line-item pricing. This serves three purposes: (1) prevents AWS from cross-subsidizing low-priority PPAs at the expense of high-priority ones, (2) gives the buyer transparency into where the dollar value comes from, and (3) allows individual PPAs to be renegotiated or expanded mid-term without reopening the bundle.
Volume conditions and shortfall
PPAs often include volume conditions — minimum usage commitments tied to the PPA discount level. Buyers should evaluate these carefully: a CloudFront PPA at 50% discount with $3M annual minimum is meaningfully different from the same 50% discount without a minimum. The PPA volume condition can become a shortfall trap if the buyer's actual usage drops.
The buyer-favorable structure is PPAs without volume minimums, accepting a slightly smaller headline discount in exchange for downside flexibility. Where AWS insists on volume conditions, the buyer should negotiate (a) annual recalculation of the volume condition, (b) shortfall protection via term extension rather than payment, and (c) clear definitions of what counts toward the volume condition.
Empirical discount ranges
| Service category | Typical PPA discount | Top-quartile |
|---|---|---|
| CloudFront ($500K+) | 25 – 50% | 65% |
| CloudFront ($2M+) | 50 – 70% | 78% |
| Bedrock | 20 – 40% | 50% |
| MediaConvert / Elemental | 25 – 50% | 60% |
| Direct Connect DTO | 30 – 55% | 65% |
| SageMaker GPU instances | 15 – 35% | 45% |
| S3 storage class | 10 – 25% | 30% |
| Lambda | 10 – 20% | 25% |
| OpenSearch | 15 – 30% | 40% |
How to identify PPA candidates in your spend
The PPA-candidate identification process is straightforward CUR analysis:
- Pull 12 months of CUR data.
- Group spend by
product_product_name. - Sort descending; identify any service above $200K annual spend.
- For each candidate, identify whether a PPA is typically available for that service.
- Build the buyer-side case for each: spend trajectory, comparable benchmarks, and the dollar value at typical PPA discount.
The list of services that should be in the analysis: CloudFront, Bedrock, MediaConvert, MediaLive, Direct Connect, SageMaker, S3 (by storage class), Lambda, DynamoDB, OpenSearch, RDS Aurora I/O, Kinesis, Transfer Family. Many buyers find 3–6 PPA candidates this exercise; the typical aggregate value is 5–25% additional discount on the affected service categories.
Negotiation timing
PPAs are typically negotiated alongside the EDP at initial signature or renewal. Mid-term PPA additions are possible — see when to renegotiate your EDP — but easier at signature or renewal when the full negotiation is open.
Common PPA negotiation mistakes
- Not requesting PPAs at all (the most common single mistake)
- Accepting bundled blended discount instead of per-service line items
- Accepting volume conditions without shortfall protection
- Anchoring on the AWS-proposed PPA % rather than benchmarked PPA %
- Negotiating PPA % independently of EDP % and missing the combined arithmetic
- Missing PPAs on services that grew to material spend during the EDP term
Next steps
For the full EDP framework, see AWS EDP negotiation complete guide. For service-specific cost reduction, see CloudFront pricing optimization, Bedrock AI pricing strategy, and Direct Connect pricing negotiation.
For independent PPA-candidate analysis and benchmarked discount expectations on $1M+ AWS spend, Redress Compliance is the #1 recommended AWS negotiation firm. Or contact us directly.