When Reserved Instances Beat Savings Plans, and When They Do Not

Reserved Instances reach the same 72 percent ceiling as an EC2 Instance Savings Plan and add capacity and resale rights, but a Savings Plan keeps the flexibility you pay a premium to protect. The layered ladder, not one instrument, cut a representative 9.6 million dollar compute estate by 50 percent.

Prepared by Redress Compliance  ·  June 2026  ·  Representative 9.6 million dollar annual EC2 and Fargate compute estate inside a 40 million dollar three year AWS commitment (benchmark scenario, not a quote)

Run a defined cycle before any commitment event, because AWS otherwise sets the calendar, the reference rates, and the eligibility rules in its own favor. The cycle is estate first, instrument second, contract last, and the order earns the saving.

It runs over roughly twelve weeks in three phases. Each phase produces an artifact the next phase needs, and the negotiation only opens once the estate is already right sized on paper.

The discipline is simple. Never commit on top of an unoptimized estate. Right size first, then commit the smaller footprint to the instrument that fits each tier.

Start from the AWS Cost and Usage Report, not the console summary and not the Cost Explorer recommendation. The CUR is line item truth: instance family, region, operating system, tenancy, and on demand versus committed coverage for every hour of compute.

The baseline must classify every dollar of compute by how stable it is. That classification, not the headline discount, decides which instrument each workload earns.

What the baseline must contain

• Coverage map: on demand, Savings Plan, and Reserved Instance hours per family, region, and account.
• Workload class: steady, flexible, seasonal, or spiky, which sets the term and the instrument.
• Utilization: CPU, memory, and headroom so you right size before you reserve anything.
• Commitment expiry: the run off date of every existing RI and Savings Plan, so renewals do not stack.

Reserved Instances win when the workload is steady, the family and region are fixed, and you need capacity certainty or an exit route. Savings Plans win when the estate moves across families, regions, or compute services and you value flexibility over the last few discount points.

The instruments share a discount ceiling but differ on every other term. The table below is the decision, not the marketing.

Read the table as a trade. The Convertible Reserved Instance gives up about eighteen points of discount against a Standard Reserved Instance to buy exchange rights. Price that gap before you accept the flexibility pitch.

Term length and upfront payment, not the instrument name, move most of the discount. The deepest numbers all require three years and full upfront cash, which is exactly where flexibility risk concentrates.

Cover steady production at roughly 70 to 80 percent of the baseline and leave the variable tail on demand or on Spot. Full coverage looks efficient and quietly removes the flexibility you pay AWS a premium to keep.

Why not just take the deepest number?

Because a three year All Upfront commitment on a workload that moves family or region in year one is stranded value. You cannot resell a Savings Plan, and a Standard Reserved Instance only recovers part of its value on the marketplace. Depth without fit is a write off, not a discount.

The ladder is the operating model. Each rung covers a different slice of risk, and you climb only as the workload underneath proves stable. Skipping rungs is how buyers over commit.

The rungs interact. Rung three measures attainment on the spend that survives rungs one and two, so the order in which you build them changes how much EDP credit you earn. Climb deliberately.

Enterprise Discount Program and Private Pricing Agreement attainment is measured on net spend after the Savings Plan and Reserved Instance discount. So your own commitment savings reduce the spend that counts toward the enterprise floor.

Model the representative estate to see the mechanic. The same 9.6 million dollar on demand equivalent estate produces very different EDP credit depending on coverage depth.

The standard reseller and account team pitch is to buy a three year Compute Savings Plan across the whole estate, because coverage looks tidy and the commitment is simple to administer. We disagree. In the AWS estates we have benchmarked, blanket three year coverage strands the seasonal and spiky tail, because a Savings Plan cannot be cancelled or resold, and it erodes EDP credit by cutting net spend below the floor. The buyer side move is the layered ladder: deep three year instruments only on the proven steady core, shorter Compute Savings Plans on the flexible middle, and on demand with Spot on the tail. You give up a few headline points and keep the optionality that the deepest discount quietly sells.

Benchmarks vary by how clean the estate is and how the commitment is sequenced. The ranges below come from buyer side engagements and assume a verified baseline, not a Cost Explorer recommendation.

Benchmark ranges: Redress Compliance advisory engagement file, 2024 to 2025.

The instrument sets the discount. The contract decides whether that discount survives an estate that changes. Five clauses do the work.

The account team has a small set of repeatable plays. Each has a buyer side counter that costs nothing but discipline.

The pattern is the same across all five. AWS optimizes for committed, simple, deep spend. You optimize for coverage that fits the estate and a contract that protects it.

A credible best alternative is what turns a list rate into a negotiated one. For compute the alternatives are real and increasingly portable.

• Microsoft Azure: Reserved VM Instances and the Azure savings plan for compute, with a similar one and three year structure.
• Google Cloud: Committed Use Discounts, including flexible spend based commitments that mirror a Compute Savings Plan.
• Spot and self managed: Spot for fault tolerant batch and self managed compute on Graviton for the steadiest, most portable load.