The Instruments and Their Discount Tiers
AWS offers four main commitment instruments, and the right AWS Savings Plans and Reserved Instances portfolio uses more than one. Compute Savings Plans commit you to a steady dollar-per-hour of compute and deliver up to 66% off On-Demand, with full flexibility across instance family, size, region, operating system, and even across EC2, Fargate, and Lambda. EC2 Instance Savings Plans and Standard Reserved Instances reach up to 72% off but tie the discount to a specific instance family or configuration. Convertible RIs sit between the two — roughly the 66% range with limited exchange rights.
The implication for portfolio design is straightforward: the extra 6 percentage points of discount on the EC2 Instance and Standard RI products is the price AWS pays you for giving up flexibility. Reserve those instruments for workloads you are genuinely confident will run unchanged for the full term, and use the flexible Compute Savings Plan as the base layer everywhere else. This is the same flexibility-versus-discount trade that runs through our EDP vs Private Pricing Agreement analysis.
| Instrument | Max discount | Flexibility | Best for |
|---|---|---|---|
| Compute Savings Plan | Up to 66% | Highest — family, size, region, EC2/Fargate/Lambda | Base layer across all compute |
| EC2 Instance Savings Plan | Up to 72% | Locked to a family in a region | High-volume, proven families |
| Standard Reserved Instance | Up to 72% | Locked configuration | Stable, predictable workloads |
| Convertible Reserved Instance | ~66% | Exchangeable for equal/greater value | Workloads that may shift family |
Coverage: Commit to the Floor
The most common and most expensive sizing error is committing to peak or average usage rather than the floor. A defensible starting point is roughly 70% of your minimum 90-day baseline hourly spend — the floor captures the lowest points in your usage curve, and a 70% commitment against it secures meaningful savings while leaving a buffer for right-sizing and demand shifts. Coverage above 85–90% of total usage is where the risk profile flips: any workload decline leaves you paying for committed compute you can no longer consume.
Commitments are leverage in the wrong direction once usage falls below them. The discount only pays off on compute you actually run — over-coverage converts a saving into a stranded liability you carry for one to three years.
Laddering: Beating the Renewal Cliff
Laddering — buying commitment in staggered tranches with deliberately offset expiry dates — is what separates a managed portfolio from a single oversized bet. If every plan expires on the same day, you hit a cliff: full On-Demand rates for however long it takes to analyse and repurchase. With offset expirations you are always partially covered, only one tranche is ever exposed, and you re-rate each rung to current usage as it matures. For a portfolio in the $500K–$2M per month range, a typical structure is three to five simultaneous plans — one or two Compute Savings Plans with staggered expiries, one Database Savings Plan, and optionally one EC2 Instance Savings Plan for a high-volume family. This same staggering logic underpins the ramp deals described in our multi-year cloud discount structures guide.
Term and Payment Trade-offs
Term and payment choice move the discount materially. The gap between a 1-year and a 3-year commitment runs 12 to 22 percentage points depending on payment option — a 1-year no-upfront plan delivers around 11%, while a 3-year no-upfront plan reaches roughly 18.5%, and all-upfront 3-year Compute Savings Plans top out near 66%. Within a given term, all-upfront beats partial-upfront by only three to five points, and partial beats no-upfront by a similar margin — so paying cash upfront buys less than committing to the longer term does.
The practical rule: lengthen the term on your stable floor (3-year, all- or partial-upfront for the deepest discount) and keep the variable top layer short and flexible (1-year, no-upfront) so it can absorb usage changes without locking you in. Mixing terms across the ladder is how you hold a high blended discount without a single large expiry exposure.
How the Portfolio Stacks With an EDP
Savings Plans and Reserved Instances are automated, self-service discounts. An Enterprise Discount Program is a negotiated commercial agreement — a different mechanism entirely. The two are complementary: the EDP sets a platform-wide discount floor across your whole bill, and the laddered RI and Savings Plan portfolio captures the deeper compute-specific discounts on top. Run both. Anchoring your strategy on the EDP alone leaves compute savings on the table; relying only on Savings Plans forgoes the cross-service discount an EDP delivers. One constraint to note: since January 2024, EDP customers can no longer sell discounted Reserved Instances on the Marketplace, which removes a former escape hatch for over-bought RIs — another reason to size to the floor. For how the negotiated layer is built, see the AWS enterprise agreement negotiation guide and the broader cloud contract complete guide.
The Overcommitment Traps
Three patterns recur. The first is buying one large 3-year all-upfront Savings Plan to maximise the headline discount, then watching a re-platforming or workload migration strand a third of it. The second is letting coverage drift toward 100% because the savings dashboard rewards it — until a seasonal trough turns the commitment into dead weight. The third is ignoring the new Database and SageMaker Savings Plans and over-buying generic Compute coverage that the more specific plans would have discounted further. A portfolio that is laddered, floored at the baseline, and split across terms avoids all three. If you want an independent review of your commitment portfolio before your next renewal, request a confidential briefing, explore the AWS vendor intelligence hub, or download the AWS EDP Negotiation Playbook.