AWS Reserved Instances and Savings Plans Optimisation: A Complete Enterprise Guide

Why Commitment Discounts Are the Most Mismanaged Cost Lever in AWS

AWS charges on-demand prices by default. For any sustained, predictable workload, paying on-demand means overpaying by 40 to 75 percent compared to rates available through commitment mechanisms. Yet across the enterprise accounts we advise, commitment coverage routinely sits below 60 percent of eligible spend — leaving substantial savings uncaptured.

The reason is not ignorance of the mechanisms. Most cloud teams know that Reserved Instances and Savings Plans exist. The problem is the operational discipline required to size, purchase, track, and renew commitments accurately as workloads evolve. Enterprises that treat RI and Savings Plans management as a one-time annual exercise consistently underperform those that operate it as a continuous programme.

This guide provides the framework for building that programme, starting with the mechanics of each instrument and ending with a practical blended strategy.

Reserved Instances: Mechanics and Types

A Reserved Instance is a commitment to use a specific EC2 instance configuration — defined by instance family, size, operating system, tenancy, and optionally region or Availability Zone — for either one year or three years. In exchange for that commitment, AWS charges a discounted hourly rate compared to on-demand pricing.

Standard Reserved Instances

Standard RIs deliver the deepest discounts, up to 75 percent below on-demand for a three-year all-upfront commitment. The trade-off is rigidity: the instance family, size, operating system, and tenancy are fixed for the commitment term. Standard RIs can be sold on the AWS Marketplace if the workload changes, but doing so adds operational complexity and may not recover full remaining value.

Standard RIs are appropriate for baseline workloads that have been stable for at least twelve months and are expected to remain stable through the commitment term. Production database servers, application servers running a core platform, and dedicated infrastructure for compliance workloads are typical candidates.

Convertible Reserved Instances

Convertible RIs sacrifice a small portion of the Standard RI discount — typically three to eight percentage points — in exchange for the ability to exchange the commitment for a different instance family, size, operating system, or tenancy at any point during the term. This flexibility makes Convertible RIs appropriate for environments where instance type modernisation is expected, such as workloads likely to migrate from older generation instances to current generation equivalents during the commitment period.

Convertible RIs cannot be sold on the Marketplace, which means unwanted commitments can only be exchanged rather than liquidated. Sizing accuracy therefore remains important even with Convertible RIs.

Regional vs Zonal Reserved Instances

Regional RIs apply a discount to any instance matching the instance family and size within the specified region, regardless of which Availability Zone the instance runs in. They also provide instance size flexibility within the same family (for example, a Regional RI for m5.xlarge provides a discount that can be applied to two m5.large instances). Regional RIs are the standard choice for most enterprise workloads because they provide automatic capacity flexibility across AZs.

Zonal RIs apply to a specific AZ and provide a capacity reservation — a guarantee that the capacity will be available when needed. This is relevant for latency-sensitive workloads where a specific AZ placement is required, or for disaster recovery scenarios requiring guaranteed capacity in a particular location.

Savings Plans: Mechanics and Types

Savings Plans were introduced by AWS in 2019 as a more flexible alternative to Reserved Instances. Instead of committing to a specific instance type, a Savings Plan commits to a minimum hourly spend in dollars — for example, $10 per hour — in exchange for discounted rates across eligible services.

Compute Savings Plans

Compute Savings Plans provide the broadest coverage: discounts apply automatically to any EC2 instance regardless of family, size, AZ, region, operating system, or tenancy, and also extend to AWS Fargate and AWS Lambda. The discount ceiling is lower than Standard RIs — up to 66 percent versus 75 percent — but the coverage breadth makes them the default choice for most enterprise workloads.

A Compute Savings Plan set at $50 per hour will apply its discounted rate to the first $50 of eligible spend per hour, across EC2, Fargate, and Lambda, in any region and with any instance type. This automatic coverage across services and regions eliminates the management overhead of tracking which RIs apply to which instances.

EC2 Instance Savings Plans

EC2 Instance Savings Plans commit to a specific instance family within a specific region (for example, m5 instances in us-east-1) but allow flexibility on instance size, AZ, operating system, and tenancy within that family and region. Discounts reach up to 72 percent — between Compute Savings Plans and Standard RIs in savings depth, with moderate flexibility.

EC2 Instance Savings Plans are appropriate when an instance family commitment can be made for a region but flexibility on size or OS is required. Environments running a mix of instance sizes within a consistent family — for example, an m5 fleet scaling from medium to 2xlarge depending on workload — benefit from this structure.

SageMaker and Database Savings Plans

AWS has extended the Savings Plans model to cover SageMaker (up to 64 percent savings) and databases including RDS, Aurora, Redshift, and DynamoDB (up to 35 percent savings). Database Savings Plans apply regardless of engine, instance family, size, deployment option, or region, making them the most straightforward commitment instrument for database spend.

Reserved Instances vs Savings Plans: The Practical Choice

The difference in maximum discount between Standard RIs (75 percent) and Compute Savings Plans (66 percent) is real but narrow — approximately three to nine percentage points depending on instance type and term. For most enterprises, the operational flexibility of Savings Plans outweighs this marginal discount advantage in almost every scenario.

The case for retaining Standard RIs is strongest where the following conditions apply: the workload is genuinely stable and unlikely to require instance changes during the commitment term; the specific instance configuration can be identified with confidence; and the RI management overhead can be absorbed into existing FinOps operations. Reserved Instances also retain the capacity reservation option through Zonal RIs, which Savings Plans do not provide.

In practice, the recommended enterprise approach in 2025 is to lead with Compute Savings Plans for coverage breadth, supplement with EC2 Instance Savings Plans for workloads with consistent family usage, and retain Standard RIs only for specific stable baselines where the marginal discount justifies the management overhead.

Sizing Commitments Correctly

Commitment waste is the core operational failure mode. Purchasing more commitment than you use means paying a discounted rate for capacity that is never consumed — a loss relative to simply not purchasing the commitment. Achieving RI utilisation above 80 percent and Savings Plans coverage above 70 percent of eligible spend are standard FinOps targets.

Rightsizing Before Committing

The single most important pre-commitment action is rightsizing current instances. AWS Cost Explorer's rightsizing recommendations identify instances consistently running at low utilisation — typically below 40 percent CPU — that could be downsized without performance impact. Purchasing an RI or Savings Plan commitment for an oversized instance locks in the cost of that oversize for one to three years. Rightsizing first, then committing to the reduced footprint, is the correct sequence.

Commitment Term Selection

Three-year commitments deliver approximately 20 to 25 percentage points more savings than one-year equivalents for Standard RIs. However, three-year terms carry proportionally higher risk of commitment waste if workloads change. Given the pace of AWS service evolution and enterprise modernisation programmes, one-year commitments are the default recommendation for most enterprise workloads. Three-year commitments are appropriate only where a workload has a demonstrated three-year stability horizon and the additional discount materially justifies the commitment.

Payment Options

AWS offers three payment options for RIs and Savings Plans: all-upfront (deepest discount), partial-upfront, and no-upfront. The all-upfront option typically provides an additional two to five percentage point discount over no-upfront. For organisations with available capital and sufficient confidence in utilisation, all-upfront one-year commitments often provide the best risk-adjusted return. No-upfront options preserve cash flexibility at a modest cost premium and are appropriate when commitment confidence is lower.

Data Egress: The Most Common Surprise Cost

RI and Savings Plans optimisation addresses compute spend — the largest component of most AWS bills. But data egress costs are the most common source of bill shock that enterprises fail to anticipate, account for 6 to 12 percent of typical cloud bills, and are not covered by any commitment instrument.

AWS charges $0.09 per GB for data transferred out of AWS to the internet from most regions. Cross-region data transfer carries additional charges of $0.02 to $0.09 per GB depending on the regions involved. Cross-AZ data transfer within the same region costs $0.01 per GB in each direction — a charge that accumulates rapidly in microservices architectures with frequent inter-service communication across Availability Zones.

Enterprise egress costs regularly reach $50,000 to $200,000 per month in large deployments, yet they are often invisible during architecture design because transfer charges do not appear in capacity planning models. Egress cost analysis should be a mandatory component of any AWS cost optimisation programme, separate from and in addition to commitment discount optimisation.

AWS has introduced egress fee waivers for customers migrating data off AWS to another cloud provider or on-premises, responding to regulatory pressure around data portability. However, these waivers apply only to migration scenarios and do not address ongoing egress costs from normal operations.

Enterprise Discount Programmes: EDP and PPA

For organisations with substantial AWS commitment, the Enterprise Discount Programme (EDP) provides contractual discounts in addition to RI and Savings Plans rates. Meaningful EDP discounts typically begin at approximately $2 million or more in annual committed spend. Below this threshold, the administrative overhead of negotiating an EDP rarely delivers material benefit over optimised commitment purchasing.

EDP discounts are applied on top of on-demand rates before Savings Plans discounts are calculated. This means EDP and Savings Plans stack — an organisation with an EDP discount of 10 percent and a Compute Savings Plan receiving a 66 percent discount applies both, achieving an effective discount of approximately 70 percent off standard list pricing.

Negotiating an EDP requires demonstrating credible committed spend forecasts, multi-year growth projections, and willingness to commit to minimum spend thresholds. Organisations that approach EDP negotiations without independent benchmarking data routinely leave five to fifteen percentage points of additional discount on the table.

Building a Continuous Commitment Optimisation Programme

One-time RI purchases are insufficient. Effective enterprise AWS commitment management requires a continuous programme with the following components:

• Monthly coverage review: Track the percentage of eligible compute spend covered by Savings Plans or RIs. Industry benchmark is above 70 percent coverage with above 80 percent utilisation.
• Quarterly rightsizing: Review AWS Cost Explorer rightsizing recommendations and apply instance downsizing before commitment renewals.
• Expiry tracking: Maintain a register of all RI and Savings Plans commitments with expiry dates. Commitments that expire without renewal revert to on-demand pricing, creating cost spikes that are frequently not caught until the following month's bill.
• Effective Savings Rate (ESR) monitoring: Measure ESR — the ratio of actual savings achieved versus the maximum achievable savings at full commitment — as the primary performance metric. Traditional coverage and utilisation metrics alone can be misleading; ESR captures the true financial performance of the commitment portfolio.
• Egress cost dashboards: Maintain separate dashboards for data transfer costs by service, region, and transfer type. Egress anomalies surface architectural inefficiencies and unexpected data movement patterns that require remediation.

Commitment Strategy for Containerised and Serverless Workloads

The shift from EC2-centric architectures to Fargate containers and Lambda serverless functions requires a corresponding shift in commitment strategy. EC2 RIs do not apply to Fargate or Lambda spend. Compute Savings Plans cover all three services, making them the essential commitment instrument for organisations with mixed EC2, Fargate, and Lambda footprints.

For Fargate-heavy environments, the commitment calculation changes: Fargate pricing is per vCPU-hour and per GB-memory-hour, which is less intuitive to model than EC2 instance hours. AWS Cost Explorer provides Fargate Savings Plans coverage recommendations based on actual usage patterns, which should be used as the sizing basis rather than manual estimation.

Lambda functions are charged per invocation and per GB-second of execution duration. Compute Savings Plans apply a discount to Lambda compute costs at the $0.0000166667 per GB-second rate. For organisations running Lambda at scale, ensuring Savings Plans coverage includes Lambda spend is important — it is frequently omitted from commitment calculations because Lambda costs appear trivial at small scale but accumulate significantly in production event-driven architectures.

Organisations that want independent analysis and negotiation support for AWS Marketplace strategy, EDP structuring, and procurement optimisation work with our AWS contract negotiation specialists. Redress Compliance is 100% buyer-side — no vendor commissions, no referral fees.

Key Takeaways for Enterprise AWS Buyers

• Compute Savings Plans should be the default for most enterprise workloads due to breadth of coverage across EC2, Fargate, and Lambda with no management overhead for instance type changes.
• Reserved Instances remain relevant for stable EC2 baselines where the marginal discount versus Savings Plans justifies the rigidity, and for zonal capacity reservations.
• Rightsize before committing. Purchasing commitments for oversized instances locks in waste for one to three years.
• Data egress is not covered by any commitment instrument and is the most common surprise cost in enterprise AWS bills. Treat egress analysis as a parallel programme.
• EDP negotiations only deliver meaningful value at $2M or more annual committed spend. Below that threshold, focus on Savings Plans and RI optimisation.
• Manage commitments continuously through monthly coverage reviews, quarterly rightsizing, and expiry tracking. Annual reviews are insufficient for dynamic enterprise environments.