AWS Savings Plans are a core component of AWS FinOps, offering discounts on compute usage in exchange for a commitment to a consistent amount of usage (measured in dollars per hour) over a one- or three-year term.
Rather than tying the commitment to specific instance types or regions, Savings Plans apply automatically to eligible AWS compute resources. This structure allows organizations to reduce costs across services like Amazon EC2, AWS Fargate, and AWS Lambda without having to make rigid, upfront instance specifications.
Savings Plans are suited for organizations that prioritize cost savings but still need the ability to adapt to changing compute requirements. The pricing model automatically applies the applicable discount to any usage matching the plan's scope, with standard on-demand rates being charged for any usage beyond the committed amount.
The flexibility of Savings Plans makes them attractive for dynamic workloads where usage may fluctuate, as users can shift resources within families, sizes, and locations without losing the benefit of discounted pricing.
AWS Reserved Instances (RIs) are a cost-saving model for Amazon EC2 where customers commit to using a specific instance type, size, and region for a one- or three-year period. This commitment grants significant discounts compared to on-demand pricing.
RIs lock customers into specific resource parameters, ensuring they receive the discount only when those exact parameters are met during usage. Their rigid structure means customers must plan their cloud infrastructure needs with accuracy, as changes to instance types, regions, or operating systems usually aren’t permitted unless using Convertible RIs.
RIs are ideal for steady-state workloads where future compute requirements can be forecasted with confidence.
Editor’s note: Updated differences between AWS discount options to reflect Amazon policy and pricing in 2026.
This is part of a series of articles about AWS cost management
Compute Savings Plans provide the most flexibility of the Savings Plans options. With Compute Savings Plans, users commit to a certain amount of compute spend per hour but aren’t restricted to specific instance families, sizes, operating systems, or regions. This means workloads can move between different EC2 instance families, across different AWS regions, and even to other eligible compute services like AWS Fargate or Lambda, all while maintaining the committed discount.
This broad applicability makes Compute Savings Plans ideal for organizations that regularly shift workloads or develop and deploy new compute resources in response to changing demand. The automatic application of discounts across services and regions removes many of the long-term planning challenges faced with traditional reserved pricing. However, discount rates are lower than other plans due to this added flexibility.
|
Term |
Payment Option |
Discount Range Compared to On-Demand |
|
1 Year |
All Upfront |
29% to 49% |
|
1 Year |
Partial Upfront |
27% to 47% |
|
1 Year |
No Upfront |
26% to 46% |
|
3 Years |
All Upfront |
42% to 66% |
|
3 Years |
Partial Upfront |
40% to 64% |
|
3 Years |
No Upfront |
38% to 62% |
Discounts vary depending on instance type, region, and utilization patterns.
EC2 Instance Savings Plans offer lower rates compared to Compute Savings Plans but require more specific commitments. When purchasing this Savings Plan, users must commit to a particular EC2 instance family in a chosen region, though they can vary instance size, operating system, and tenancy within that family. This model applies discounts only to usage matching these commitments.
These plans suit organizations with stable workloads running on a specific EC2 instance family. While less flexible than Compute Savings Plans, the discounts are generally higher due to the additional restrictions.
|
Term |
Payment Option |
Discount Range Compared to On-Demand |
|
1 Year |
All Upfront |
31% to 52% |
|
1 Year |
Partial Upfront |
30% to 50% |
|
1 Year |
No Upfront |
28% to 48% |
|
3 Years |
All Upfront |
48% to 72% |
|
3 Years |
Partial Upfront |
46% to 70% |
|
3 Years |
No Upfront |
44% to 68% |
Discounts are higher due to restrictions on instance family and region.
SageMaker Savings Plans focus exclusively on AWS SageMaker, the company’s managed machine learning service. Customers commit to a stated dollar-per-hour spend on SageMaker usage, covering various SageMaker features like training jobs, real-time inferencing, batch transform jobs, and processing.
The discounts automatically apply to any eligible SageMaker usage, regardless of instance family or region within SageMaker.
|
Term |
Payment Option |
Discount Range Compared to On-Demand |
|
1 Year |
All Upfront |
24% to 30% |
|
1 Year |
Partial Upfront |
22% to 28% |
|
1 Year |
No Upfront |
20% to 26% |
|
3 Years |
All Upfront |
34% to 64% |
|
3 Years |
Partial Upfront |
32% to 62% |
|
3 Years |
No Upfront |
30% to 60% |
Coverage includes SageMaker training, inference, and processing features.
Standard Reserved Instances provide the deepest discounts among Reserved Instance types—up to 72% compared to on-demand pricing—for customers willing to commit to a specific EC2 instance type, size, platform, and region.
This type of RI is strictly non-flexible, so any changes to instance attributes typically mean the discount no longer applies, requiring users to plan their capacity needs with accuracy.
The main benefit is cost efficiency for static workloads, but the drawback is the lack of flexibility. If usage patterns change, organizations may find themselves unable to adjust without incurring additional costs outside their RI allocation.
Convertible Reserved Instances offer more flexibility than Standard RIs. They allow customers to exchange their RI for another of equal or greater value, as long as the new reservations are for EC2 instances. This feature makes Convertible RIs better suited for workloads where future needs might be uncertain or subject to change. Discount rates for Convertible RIs are lower than those of Standard RIs but remain significant over on-demand prices.
Convertible RIs are valuable for cloud strategies that anticipate change but still need discounted, predictable billing. Teams can adapt instance families, operating systems, or tenancy during the term of their RI without losing reservation benefits.
|
Term |
Payment Option |
Discount Range Compared to On-Demand |
|
1 Year |
All Upfront |
27% to 41% |
|
1 Year |
Partial Upfront |
25% to 39% |
|
1 Year |
No Upfront |
23% to 37% |
|
3 Years |
All Upfront |
45% to 54% |
|
3 Years |
Partial Upfront |
43% to 52% |
|
3 Years |
No Upfront |
41% to 50% |
Both AWS Savings Plans and Reserved Instances are designed to reduce costs for long-term cloud compute usage by requiring a commitment from customers. They offer discounts compared to on-demand pricing.
In both models, customers commit to usage over a one- or three-year term. This commitment enables AWS to offer lower prices, since it aligns infrastructure provisioning with customer demand. Whether through a dollar-per-hour commitment (Savings Plans) or instance-specific reservations (RIs), the billing structure rewards long-term planning.
Additionally, both Savings Plans and RIs are applied automatically to eligible usage. AWS handles the application of these discounts behind the scenes, reducing administrative overhead and ensuring optimal utilization of committed resources.
AWS Savings Plans and Reserved Instances both offer meaningful cost reductions compared with On-Demand pricing, but they differ slightly in where they shine. With Savings Plans, you commit to a fixed dollar-per-hour spend over a one- or three-year term, and AWS applies discounted pricing automatically to eligible compute usage.
Compute Savings Plans typically provide savings of up to about 66%, while EC2 Instance Savings Plans can go up to roughly 72% off On-Demand rates, depending on the plan type and payment option you choose. Because Savings Plans cover usage across multiple compute services such as EC2, Lambda, and Fargate, they can deliver deep savings for varied workloads.
Reserved Instances (RIs) also provide significant discounts, often up to 72–75% off On-Demand costs, especially for Standard RIs with all-upfront payment. Standard RIs are typically slightly better in pure discount potential when you can lock into specific instance families, regions, and terms, making them optimal for predictable, long-running compute usage. Meanwhile, Convertible RIs offer somewhat lower maximum savings but allow limited reconfiguration later, blending savings with flexibility.
One of the key operational differences between the two models lies in capacity guarantees. Savings Plans do not reserve specific compute capacity, they purely offer price discounts based on your committed spend and usage.
If you require guaranteed compute capacity in a particular Availability Zone (AZ), such as for mission-critical workloads, you must provision On-Demand Capacity Reservations separately; Savings Plans still apply discounts to that usage. This means Savings Plans are best for cost savings rather than guaranteeing resource availability.
In contrast, Reserved Instances, particularly Zonal RIs, can reserve actual compute capacity within a chosen AZ. This capacity reservation is valuable for applications that must avoid capacity shortages, for example, production databases or services with strict uptime requirements. Regional RIs don’t lock capacity but still ensure the discount applies across the entire region to matching instance usage.
Savings Plans are designed around flexibility in both usage and pricing application. With a compute Savings Plan, the discount automatically applies to usage across any EC2 instance family, OS, region, size, tenancy, and even extends to serverless compute services like AWS Lambda and AWS Fargate.
Reserved Instances are more configuration-specific. Standard RIs require you to specify instance family, region, OS, and other attributes, and the associated discount only applies if your running instances match those characteristics.
Convertible RIs provide some flexibility by allowing you to exchange the RI for another with different attributes (as long as value is equal or greater), but this still involves manual exchange actions and planning. This means RIs are less flexible than Savings Plans, though they are often a good fit when workloads are stable and well-understood.
Both Savings Plans and Reserved Instances require a commitment for 1-3 years, but they handle growth and change differently after purchase. Savings Plans are immutable after purchase beyond a brief AWS refund period (typically around a week).
Once committed to a specific spend amount, you cannot adjust that commitment upward or modify the plan; if your needs grow, you must buy additional Savings Plans. This straightforward model reduces management overhead but means growth requires incremental purchases rather than plan edits.
With Reserved Instances, Convertible RIs allow more configurational flexibility over time. You can exchange Convertible RIs for different instance families, sizes, or OS configurations as long as the new RI has equal or greater value, which lets you adapt your reservation to evolving workloads without losing the original commitment entirely.
Standard RIs are more restrictive and are best when usage patterns are well-known and unlikely to change, as they cannot be reconfigured after purchase.
How you can sell or trade your commitment is another important difference. Savings Plans, once purchased and beyond the short cancellation window, cannot be resold, exchanged, or transferred to other parties. You’re committed to your spend until the plan expires, and AWS doesn’t provide a marketplace or exchange mechanism for Savings Plans.
Reserved Instances, however, do offer secondary market options, at least for certain types. Standard RIs that you no longer need can often be listed on the AWS Reserved Instance Marketplace, allowing other AWS customers to purchase them and recoup some of your investment.
Also, Convertible RIs can be internally exchanged for new configurations mid-term (subject to value-matching rules), providing a degree of lifecycle flexibility not available with Savings Plans.
Choosing between AWS Savings Plans and Reserved Instances depends on your organization’s workload predictability, flexibility needs, and commitment strategy:
Use Savings Plans when flexibility and service coverage are priorities, and opt for Reserved Instances when deep discounts for consistent, static usage are the goal. Organizations with mixed workloads may even benefit from a blended strategy—combining Savings Plans for variable compute and Reserved Instances for predictable baselines.
Finout's FinOps solution is particularly adept at managing AWS costs, including those associated with Amazon Elastic Kubernetes Service (EKS), making it an excellent tool for organizations leveraging AWS's extensive cloud services. It facilitates real-time cost allocation and reassignment across the entire AWS infrastructure, which is pivotal for companies with intricate and dynamic cloud environments.