EDPs and Savings Plans stack: AWS applies both to your usage. If your EDP gives 25% off EC2 On-Demand list and your Compute Savings Plan gives 22% off the same list, AWS picks the larger of the two per instance, not the additive total. The practical implication is that a customer who already has an EDP at 25% gains almost nothing from buying a Compute SP at 22%, because the SP discount will never exceed the EDP discount on the same instance. The right pattern is to layer an EDP on top of existing RIs/SPs only when the EDP discount exceeds the average RI/SP discount by at least 5 percentage points. For most enterprises above $5M/year, EDP alone typically beats the optimal Savings Plan + On-Demand mix (AWS Pricing, AWS Savings Plans documentation, September 2026).
| Instrument | Discount mechanism | Term | Flexibility | Best fit |
|---|
| AWS EDP | Auto-applied tier | 3-5 years | Highest (all services) | $1M+/year, multi-year commit |
| Compute Savings Plan | Dollar/hour commit | 1-3 years | Any region/family/OS | $200K+/year, predictable workloads |
| EC2 Instance Savings Plan | Family+region commit | 1-3 years | Family+region locked | Steady-state, one region |
| Standard Reserved Instance | Per-instance reservation | 1-3 years | Family+region+tenancy | Legacy workloads, predictable |
| Convertible Reserved Instance | Per-instance reservation | 1-3 years | Exchange during term | Workloads that change |
Source: AWS Savings Plans, AWS Reserved Instances pricing documentation, September 2026.
EDP negotiation playbook for 2026
The most successful EDP negotiations in 2026 start 6-9 months before renewal with a credible multi-cloud threat model.
Step one is to build a 24-month forecast by AWS service family with quarterly projections: EC2 by instance family, S3 by storage class, RDS by engine, data transfer by region, and any growth services like Bedrock or SageMaker. The forecast should include both committed baseline (the floor) and projected upside (the upside tiers). AWS will challenge this forecast during the negotiation, so it needs to be defensible from real CUR data and a bottoms-up workload inventory.
Step two is to model discount tiers at 5-percentage-point increments from 15% to 45% across EC2, S3, RDS, and data transfer, and calculate the dollar value of each tier for the customer's specific service mix. A customer with $5M/year in EC2 and $1M/year in S3 sees a very different value from a 30% EC2 discount than a customer with $2M/year in EC2 and $4M/year in S3. Step three is to build a competitive posture: pull Microsoft Customer Agreement (MCA) discounts from Azure and Google Committed Use Discount (CUD) tiers for the equivalent workloads, and have those numbers ready when AWS negotiates.
Step four is to identify workloads that could move clouds in 6-12 months - typically stateless APIs, containerized microservices, and greenfield data platforms - and have an internal estimate of the migration cost. AWS responds to credible multi-cloud optionality by 5-10 percentage points of additional discount. Most EDPs are signed in 8-16 weeks of active negotiation with a named AWS enterprise account manager, with the largest deals taking 6+ months. The biggest mistake is treating EDP as a procurement event rather than a multi-quarter strategic exercise (AWS Partner negotiation playbooks, AWS re:Invent 2025, September 2026).
Common EDP pitfalls and how to avoid them
The most common EDP failure mode in 2026 is over-committing to a floor the customer cannot hit during a workload migration or business downturn.
EDPs include a minimum annual spend floor with monthly pro-ration, meaning if actual spend in any month falls below the pro-rated floor, the customer still pays the difference. Customers who over-project growth during a 3-year term - for example, signing a $5M/year floor when actual consumption is $3M/year for two of the three years - end up paying the gap, which can be 30-50% of the contractual savings. The right preparation includes a downside scenario in the forecast: what does the bill look like if AWS workloads are rightsized, a workload moves to Azure, or a business unit is divested?
The second common pitfall is signing an EDP that excludes the newest services (Bedrock, SageMaker, Lambda, GPU instances for AI workloads). If the customer is investing in generative AI, an EDP that gives 25% off EC2 but lists Bedrock and SageMaker at 0% discount can become a margin problem as AI workloads scale. The fix is to negotiate growth-tier language that extends EDP discounts to new service families as spend grows, and to insist on a rate-card update clause tied to new AWS services launched during the term (AWS Enterprise Agreement FAQ, AWS Partner negotiation playbooks, September 2026).
FAQs
Q: How long does an AWS EDP negotiation take in 2026?
A typical first-time EDP negotiation for a $5M/year customer takes 10-16 weeks from initial AWS sales engagement to signed agreement, with renewals typically taking 6-10 weeks because the customer has more leverage and historical data. Negotiations above $20M/year often span 16-24 weeks because of legal review, executive sign-off, and redlines on the master agreement. The customer should expect 4-6 weeks of pre-negotiation preparation before AWS engagement starts: build the forecast, model the discount tiers, and assemble the competitive posture. The biggest avoidable delay is waiting until the last 60 days before renewal to engage AWS; enterprise account teams typically need 90-180 days to socialize an EDP internally (AWS, September 2026).
Q: Can I combine EDP with Savings Plans and Reserved Instances?
Yes. EDP discounts apply to list-price consumption automatically, and AWS applies Savings Plans and Reserved Instance discounts on top of the EDP-discounted price. The Math: if EDP discounts EC2 On-Demand by 25% and a Compute SP commits to a $100/hour spend at 22% off, the SP discount applies on top of the EDP-discounted rate. In practice, however, the SP discount rarely exceeds the EDP discount on the same workload, so the marginal value of buying SPs after an EDP is small. The right pattern is to keep existing RIs in place during the first 12 months of an EDP and let them expire naturally, then avoid new SP/RI purchases in services where EDP already gives 25%+ off (AWS Pricing documentation, September 2026).
Q: Does EDP cover AWS Marketplace spend?
Partially. EDPs include a Marketplace clause that applies a smaller discount tier (typically 5-10%) on third-party software purchased through AWS Marketplace, but the third-party vendor sets the gross price. For customers spending heavily on Datadog, Snowflake, Confluent, or MongoDB through Marketplace, the EDP Marketplace discount is a small win; the bigger lever is to negotiate directly with the third-party vendor for an enterprise discount. Some EDPs include language that lets customers apply their own enterprise discounts to Marketplace spend by paying AWS directly and then passing through to the third-party vendor; this is a negotiation tactic reserved for $20M+ enterprise agreements (AWS, September 2026).
Q: What happens at EDP renewal if my spend has dropped 30%?
Most EDPs include a mid-term true-up clause that allows the customer to revise the annual floor by 10-20% up or down at the 12 or 18 month mark, but a 30% drop requires renegotiation. AWS's preferred path is to revise the rate card to keep the dollar value similar: if consumption drops 30%, AWS will offer 3-5 additional percentage points of discount to recover the lost revenue. Customers in this position should also evaluate whether to terminate the EDP early (50-100% of remaining commitment as exit fee) or convert to a smaller ECR/PPA. The exit math is usually unfavorable because the remaining commitment is typically larger than the savings differential (AWS, September 2026).
Q: Are EDPs available in AWS GovCloud or AWS China regions?
EDPs are available in AWS GovCloud (US) with the same structure but require FedRAMP-aligned invoicing and US-person account sponsorship. EDPs in AWS China (operated by Sinnet for Beijing and Ningxia regions, or NWCD for Ningxia) are handled separately by the Chinese partners and have different discount tiers, longer negotiation cycles, and additional regulatory review. Most multinational enterprises operate separate EDPs for US, EMEA, APAC, and China regions, with the master EDP covering the bulk of global consumption and regional addendums covering sovereign-cloud workloads (AWS GovCloud, AWS China documentation, September 2026).
Q: How do EDPs compare to Microsoft Customer Agreement and Google CUA in 2026?
AWS EDPs are more aggressive than Microsoft's MCA-EA and Google's CUA at the high end of the spend tier. Published customer reports suggest MCA-EA typically delivers 10-30% off list for $1M-$20M/year Azure commits, while Google CUA reaches 20-37% off list for similar GCP commits. AWS EDPs in the same band reach 22-35%, and at $50M+/year AWS often matches or beats Microsoft's 35-45% Enterprise Agreement tier and Google's 30-40% Premier tier. The reason is competitive intensity: the three hyperscalers price-discount against each other in every major account, which is why credible multi-cloud optionality in negotiations matters so much (AWS, Microsoft, Google Cloud pricing pages, September 2026).
Q: Can a startup under $1M/year negotiate an EDP-equivalent?
Startups in the $200K-$1M/year band have several EDP-equivalent options. The AWS Activate program offers $1K-$100K in credits depending on the funding stage but no rate-card discounts. The ECR program kicks in around $500K/year for a 1-3 year term with 5-15% off list. PPAs start as low as $200K/year for 5-12% off list. For pre-revenue startups, the right path is Activate credits plus on-demand pricing; for Seed-Series startups, PPAs; for Series A/B startups with growth, ECR; for Series C+ startups at $1M+/year, full EDP. The mistake is trying to negotiate an EDP before $1M/year - AWS's enterprise sales motion is built for $1M+ accounts and the time investment is rarely worth it for sub-$1M customers (AWS Activate, AWS Sales FAQ, September 2026).
Fazlur Rahman
