What is the difference between Compute Savings Plans and EC2 Instance Savings Plans?

Compute Savings Plans apply across EC2, Fargate, and Lambda usage regardless of instance family, region, or operating system, in exchange for a lower discount. EC2 Instance Savings Plans are family- and region-specific and offer a deeper discount in exchange for less flexibility.

How much should we commit to Savings Plans?

The right commitment level covers the steady-state compute floor — usage you are confident will exist for the full term regardless of growth scenarios. Anything above the floor should be left to on-demand or layered into shorter-term commitments.

Should we still use Reserved Instances?

In most cases Compute or EC2 Savings Plans now offer better flexibility-to-discount ratios than Standard or Convertible RIs. Exceptions remain for RDS, Redshift, ElastiCache, and OpenSearch workloads, where RIs are still the only commitment vehicle.

Compute Savings Plans vs EC2 Instance Savings Plans — which is better?

Compute Savings Plans apply across EC2, Fargate and Lambda and let you move freely between instance families and regions, in exchange for a lower discount (typically 5-10 points below EC2 SPs). EC2 Instance SPs lock you to a family and region but deliver up to 72% off. The right answer is usually a blend, with EC2 SPs covering your most stable workloads and Compute SPs absorbing the variable tail.

How much of my compute spend should be covered by Savings Plans?

For most steady workloads, 70-85% coverage with a blend of 1-year and 3-year terms maximizes savings while preserving flexibility for growth or migration. Coverage above 90% creates risk if you change architectures or move regions.

Can I convert existing Reserved Instances to Savings Plans?

Not directly — RIs cannot be converted into SPs. But you can let RIs expire, sell convertible RIs on the Marketplace, or layer SPs alongside existing RIs (SPs apply first, RIs cover the gap). The transition window is where most savings are lost or gained.

What's the right Savings Plan commitment level?

Model your steady-state hourly compute floor (the bottom of your usage curve over the last 60 days), then commit at that level. Anything above the floor risks unused commitment; anything below leaves on-demand pricing exposed on your baseline.

Are Savings Plans negotiable beyond the published rates?

The published SP rates themselves are fixed, but the effective price you pay can be further reduced through EDP overlays and Private Pricing Addenda. We routinely secure 8-12 additional discount points on Savings Plans through EDP-bound PPAs.

Service · Savings Plans

AWS Savings Plans optimization, done right.

By AWSNegotiations Practice·Published February 12, 2025·Last updated March 2, 2026·9 min read

Coverage analysis, commitment right-sizing, Compute vs EC2 plan strategy, and clean conversion from legacy Reserved Instances. We model your steady-state compute floor and structure commitments that maximize discount without trapping spend.

$2.4B+

AWS spend reviewed

500+

Engagements

38%

Avg reduction

$340M+

Documented savings

Overview

Savings Plans should compound
your EDP, not undermine it.

Savings Plans are AWS's flexible commitment vehicle for compute spend. Done well, they layer cleanly under an Enterprise Discount Programme and produce an effective discount that is materially deeper than either lever alone. Done poorly, they trap spend in commitments that survive workload migrations, organizational changes, and architectural pivots — for one to three years at a time.

Most organizations are either under-committed (paying on-demand rates on workloads that have run unchanged for two years) or over-committed (carrying Savings Plans whose coverage is no longer aligned to actual usage). Both failure modes are common, and both are expensive. The right answer is rarely one big commitment — it is a layered portfolio of overlapping commitments sized to the floor, with deliberate gaps left for workloads that may move.

We have analyzed Savings Plans portfolios across thousands of accounts and built the modeling tools to do this fast. We tell you what to commit, at what term, in what mix of Compute vs EC2 — and which RIs to let expire rather than convert.

What we optimize

Every lever in the Savings Plans portfolio.

01Steady-state floor analysis+

We model 90 days of usage data at the hourly level to identify the compute floor — the layer of usage that has been consistently present and is structurally unlikely to disappear. That floor is what you commit to. Anything above it stays flexible.

02Compute vs EC2 Instance plan mix+

Compute Savings Plans cover EC2, Fargate, and Lambda with maximum flexibility but a lower discount. EC2 Instance Savings Plans deliver deeper discounts tied to a specific instance family and region. We model the optimal mix for your workload portfolio and migration plans.

03One-year vs three-year commitment ladder+

Three-year commitments offer the deepest discount. One-year commitments preserve optionality for workloads on a migration roadmap. We build a layered ladder — three-year commitments for the rock-solid base, one-year for the moderate-confidence layer, on-demand for the uncertain top — to balance discount against optionality.

04Legacy RI conversion strategy+

Most organizations carry a tail of legacy Standard and Convertible RIs. Some should be converted, some should be sold on RI Marketplace, some should be allowed to expire. We map every RI to its optimal disposition and execute the migration without coverage gaps.

05Coverage monitoring and rebalancing+

Savings Plans are not a "set and forget" purchase. As workloads evolve, coverage drifts. We set up quarterly rebalancing cadence with explicit reinvestment triggers — when coverage drops below threshold, when new workload classes appear, when account consolidation changes attribution.

06Integration with EDP commitment+

Savings Plan spend counts toward EDP commitment but at the post-discount rate. We model this interaction explicitly so that Savings Plan purchases reinforce — rather than undermine — your EDP commitment trajectory.

Process