AWS Savings Plans vs Reserved Instances

AWS offers two primary commitment-based pricing mechanisms for reducing EC2 costs: Savings Plans and Reserved Instances. Both provide significant discounts over On-Demand pricing in exchange for a 1-year or 3-year commitment, but they differ substantially in flexibility, specificity, and management complexity. Choosing the right one -- or the right combination -- depends on your workload stability, organizational maturity, and appetite for commitment risk.

EC2 Overview

Amazon EC2 provides resizable compute capacity in the cloud. You select an instance type (defining CPU, memory, storage, and networking), launch it, and pay for the time it runs. At On-Demand rates, EC2 is straightforward but expensive. Both Savings Plans and Reserved Instances exist to reduce that per-unit cost for workloads that will run consistently over time.

What Are Savings Plans?

Savings Plans, introduced by AWS in 2019, require a commitment to a consistent dollar amount of compute usage per hour for a 1-year or 3-year term. Rather than committing to specific instances, you commit to a spend level. AWS then applies the discounted rate to your qualifying usage automatically, starting with the highest-discount usage first.

There are three types of Savings Plans:

• Compute Savings Plans: The most flexible option. Discounts apply across any EC2 instance family, size, operating system, tenancy, and region. They also cover AWS Fargate and Lambda usage. This flexibility makes them suitable for organizations with dynamic or evolving infrastructure.
• EC2 Instance Savings Plans: Scoped to a specific instance family (such as M5) in a specific region. They offer deeper discounts than Compute Savings Plans in exchange for less flexibility. Best for organizations with stable workloads concentrated in known instance families and regions.
• SageMaker Savings Plans: Specifically for Amazon SageMaker machine learning workloads. Apply to SageMaker instance usage across instance families, sizes, and regions.

Savings Plans: Pros

• Broad flexibility, especially with Compute Savings Plans -- changes in instance type, region, or operating system do not invalidate your commitment.
• Simpler to manage than Reserved Instances at scale. No need to track individual reservations across accounts.
• Automatically applies to the highest-discount usage first, optimizing your discount without manual intervention.
• Covers serverless compute (Lambda, Fargate) in addition to EC2.

Savings Plans: Cons

• Cannot be resold. If your needs change mid-term, you are locked in with no exit option.
• No capacity reservation. Your commitment guarantees a price, not instance availability.
• The "up to 72%" marketing figure is misleading. Realistic savings for most organizations are closer to 23% when accounting for workload variability and unused commitment.
• Choosing between three plan types and three payment options (All Upfront, Partial Upfront, No Upfront) adds decision complexity.

What Are Reserved Instances?

Reserved Instances (RIs) predate Savings Plans and offer discounts by committing to a specific instance type in a specific region for a 1-year or 3-year term. The commitment is to instance attributes, not a dollar amount.

Two types are available:

• Standard Reserved Instances (SRIs): Locked to a specific instance type, platform, tenancy, and region. Offer the deepest discounts but cannot be modified after purchase. Can be resold on the RI Marketplace.
• Convertible Reserved Instances (CRIs): Allow you to change the instance family, operating system, and tenancy during the term. Lower discounts than Standard RIs, but significantly more flexible. Cannot be sold on the RI Marketplace.

Reserved Instances: Pros

• Can include capacity reservation, guaranteeing that the instance type will be available when you need it -- valuable for mission-critical applications in high-demand regions.
• Standard RIs can be resold on the RI Marketplace if your needs change, providing an exit option that Savings Plans lack.
• Convertible RIs offer meaningful flexibility to adjust during the term as workloads evolve.
• Mature, well-understood pricing mechanism with extensive tooling support.

Reserved Instances: Cons

• Less flexible than Savings Plans -- Standard RIs are tied to specific instance attributes and cannot be modified.
• More complex to manage at scale, especially across multiple AWS accounts and regions.
• Requires accurate long-term forecasting of instance type, region, and capacity needs.
• Convertible RIs provide flexibility but at reduced discount levels compared to Standard RIs.

Key Differences

Flexibility

Savings Plans commit to a dollar amount and apply flexibly across instance attributes. Reserved Instances commit to specific instance configurations. For organizations with evolving infrastructure -- changing instance types, migrating regions, adopting new processor families -- Savings Plans provide significantly more protection against commitment waste. For organizations with stable, predictable workloads, the specificity of Reserved Instances can yield deeper discounts.

Usage and Capacity Reservation

Only Reserved Instances offer capacity reservation, guaranteeing instance availability in a specific Availability Zone. Savings Plans guarantee a price but not capacity. If your workload requires guaranteed availability in a specific location, Reserved Instances are the only option.

Pricing Models

Both offer All Upfront, Partial Upfront, and No Upfront payment options. The fundamental difference is what you are committing to: Savings Plans commit to an hourly dollar amount, while Reserved Instances commit to a specific instance type. This distinction matters most when your infrastructure changes -- a Savings Plan adapts automatically, while an RI may become a stranded commitment.

The 23% Reality Check

Both AWS marketing materials and many third-party guides highlight savings figures of "up to 72%." This number applies to a narrow set of conditions: 3-year term, all-upfront payment, specific instance types, and full utilization of the commitment. In practice, most organizations realize effective savings closer to 23% after accounting for unused commitments, workload variability, the cost of capital tied up in upfront payments, and the portions of their spend that do not qualify for commitment discounts. Set realistic expectations before purchasing either mechanism.

Recommendation

In practice, most organizations benefit from a layered approach:

• Use Savings Plans for the flexible, variable portion of your compute -- workloads that may change instance type, migrate regions, or evolve architecturally over the commitment term.
• Use Reserved Instances for stable, predictable workloads that will not change instance type or region, especially where capacity reservation is required for mission-critical applications.
• Layer both on top of rightsizing (to eliminate waste before committing), Spot usage (for fault-tolerant workloads), and scheduling (to eliminate idle resource costs).

The commitment decision should always follow optimization -- never commit to capacity you have not first rightsized.

Glassity automates AWS cost optimization — delivering up to 50% savings on EC2 and 69% on RDS. You only pay 10% of what we actually save you. Book a free savings assessment.