Ask the Cloud Expert: The Truth About AWS Savings Plans & AWS Reserved Instances

AWS continues to dominate the market as the biggest cloud infrastructure provider, taking the lion’s share of the market at 30%. However, having AWS as your cloud infrastructure can quickly turn out to be a costly affair, and numbers prove it too. Cloud computing, more commonly associated with AWS, accounts for around 32% of a company’s IT budget, which is a significant portion of its capital expenditures (capex).

As workloads scale, it is not feasible to provision all compute instances on demand; since on-demand instance pricing is steep, that’s where discount programs and offerings come into play. The two most popular discount programs associated with AWS are Reserved Instances(RIs) and Savings Plans, with the latter being introduced later. On the surface, AWS RIs offer up to 75% discount on certain configurations, whereas AWS Savings Plans offer up to 72% discount, based on their commitment model.

However, it’s not all a rosy picture. There are several caveats and nuances to consider regarding both AWS Reserved Instances and AWS Savings Plans.

By the end of this article, you will be empowered with real-world knowledge to know when to choose between  AWS Savings Plans Vs. Reserved Instances based on your workload and requirements, thus helping you extract maximum potential out of your AWS investment.

Featured Expert: Aman Khandelwal

Aman Khandelwal is a Senior DevOps Engineer at CloudKeeper, well-versed in the technical aspects of DevOps and experienced in managing enterprise-scale production AWS infrastructure across diverse workloads. He also possesses a strong theoretical understanding of FinOps principles. By combining deep technical expertise with FinOps knowledge, Aman helps maximize performance while reducing cloud wastage, ultimately driving higher ROI.

So, let’s get started.

Part 1: Understanding the Current Landscape of RIs and Savings Plans

Q1. What is the state of AWS Reserved Instances in 2025? Is it true that Reserved Instance discounts are being phased out in favor of AWS Savings Plans?

Reserved Instances are not being phased out in 2025. However, AWS has clearly positioned Savings Plans as the superior and more flexible discount model for the vast majority of use cases.

While AWS RIs remain available for purchase, their strategic importance has diminished significantly in favor of the simplicity and flexibility of AWS Savings Plans.

AWS RIs are still available, but not promoted: You can still buy Standard and Convertible RIs for a 1 or 3-year term. However, AWS's console, sales teams, and recommendations now almost universally steer customers toward AWS Savings Plans.

For specific use cases, AWS RIs can still be marginally better for predictable, static workloads where you are certain you will need a specific instance type in a specific availability zone for the full term. However, for most modern, dynamic environments, this is rare.

In essence, AWS is not doing away with RIs, but they are not actively pitching them to customers. For nearly all customers, Savings Plans are now the default, recommended choice because they provide the same level of discount without the operational overhead and risk of being locked into specific instance types.

Q2. What are AWS Savings Plans, and how are they different from AWS Reserved Instances?

AWS Savings Plans are flexible pricing models that allow you to commit to a specific hourly spend (measured in USD/hour) on AWS compute usage for a 1- or 3-year term. In return, you receive significant discounts—up to 72%—compared to on-demand pricing. Unlike Reserved Instances, which require committing to a specific instance type, family, or region, AWS Savings Plans apply automatically to any eligible compute usage across EC2, Fargate, and Lambda.

The key difference is flexibility. Reserved Instances lock you into particular configurations and capacity reservations, whereas AWS Savings Plans only lock in your spend commitment. This means that with Savings Plans, you can change instance types, switch regions, or even move workloads between Amazon EC2 and AWS Fargate without losing your discount.

Part 2: Discounts, Pricing Models & Commitments

Q3. What factors determine AWS Reserved Instances discounts, and which configurations receive the maximum discounts?

AWS Reserved Instance discounts aren't a flat rate; they are a complex algorithm designed to reward commitment and behaviors that benefit AWS's capacity planning. The maximum discount is a product of your willingness to accept the most restrictive terms.

Here are the key factors that determine the discount percentage:

a) Commitment Term: This is the biggest lever.

• 1-Year Term: Lower discount.
• 3-Year Term: This provides the highest possible discount. You are locking in with AWS for a longer period, and they reward that with significantly lower rates.

b) Payment Option: This affects your upfront cost and the discount rate itself, but it impacts your cash flow.

• No Upfront: You pay nothing initially and are billed a discounted hourly rate each month. The total cost is the highest.
• Partial Upfront: You pay a portion of the cost upfront and a lower hourly rate. This offers a more effective discount than No Upfront.
• All Upfront: You pay the entire cost of the term at once. This provides the highest effective discount because AWS gets all its money immediately.

c) Instance Flexibility: This is a trade-off between discount and flexibility.

• Standard RIs: These are locked to a specific instance type (e.g., m5.2xlarge), AZ, and OS. They receive the maximum discount but carry the highest risk if your needs change.
• Convertible RIs: These can be exchanged for different instance types or families later, but only in the same geographic region. They offer fewer discounts than Standard RIs because you are paying for the flexibility to change your mind.
• Region and Instance Type: Discounts may vary slightly by region due to local market conditions, supply, and demand. Newer, in-demand instance generations (e.g., Graviton) may also have different discount structures.

d) Scope:

• Regional RIs: Apply to an instance in any Availability Zone within a region. They offer a slightly lower discount than Zonal RIs but provide flexibility during outages.
• Zonal RIs: Apply to a specific Availability Zone, which helps in cost optimization. They receive a marginally higher discount but are less flexible.

The absolute maximum discount is achieved by combining the most restrictive options:

A 3-Year, All Upfront, Standard Reserved Instance, scoped to a specific Availability Zone.

This configuration gives AWS exactly what it wants: maximum upfront cash, a long-term commitment, and a guarantee of how and where it will utilize its capacity. You are accepting all the risk of your infrastructure needs changing in exchange for the lowest possible rate.

Important Caveat: While this gets the highest discount, it also carries the highest risk. Most modern organizations prioritize flexibility over maximizing every last percentage point of discount, which is why Savings Plans (which offer significant discounts with much greater flexibility) have become the default recommendation for most use cases.

Q4. What are the shortfalls of AWS Reserved Instances?

AWS Reserved Instances (RIs) are a powerful tool for savings, but they come with significant drawbacks that have led many organizations to shift towards Savings Plans. Their primary shortfalls are a lack of flexibility and high management overhead.

These are the key shortfalls of AWS RIs:

• Inflexibility and Risk of Waste: This is the biggest flaw. AWS RIs are locked to a specific instance type, size, Availability Zone, and OS. If your application changes and you no longer need that specific resource, the AWS RI becomes useless, and you are stuck paying for it. This often leads to "AWS RI waste," where companies are forced to run outdated workloads just to utilize commitments.
• High Management Overhead: AWS Reserved Instances are not "set and forget." They require constant management to ensure the best utilization. You have to track expirations, negotiate exchanges for Convertible RIs, and manually assign them to accounts in an Organization. This creates an administrative burden that often negates the value of the savings.
• Limited Scope: Traditional AWS RIs primarily only apply to EC2, RDS, and a few other compute services. They do not cover modern, flexible services like AWS Fargate or Lambda, leaving a growing portion of your bill ineligible for discounts.
• Complex Capacity Planning: You are effectively playing a guessing game with your future infrastructure needs. If you under-purchase, you leave savings on the table. If you over-purchase or guess wrong, you are financially penalized. This is incredibly difficult for dynamic or fast-changing environments.
• Upfront Financial Commitment: The payment options (All Upfront, Partial Upfront) require significant capital expenditure (CapEx), which can strain budgets and limit financial flexibility compared to the pure operational expense (OpEx) model of pay-as-you-go.

To sum up, AWS Reserved Instances require you to trade flexibility for savings. In a modern cloud environment where agility is paramount, this is often a poor trade-off. This is precisely why AWS introduced Savings Plans, which provide similar discounts but automatically apply to your dynamic usage, effectively solving the core shortfalls of the AWS RI model.

Q5. In which use cases do you not recommend Savings Plans to your customers?

While Savings Plans are the most flexible and recommended discount instrument for the vast majority of workloads, they are not a universal solution. I typically advise against them, in a few specific scenarios where their value diminishes or the commitment becomes a liability.

Here are the key use cases where Savings Plans are not a good fit:

• Unpredictable or Short-Term Workloads: For a project with a defined end date (e.g., less than 6 months) or a workload for which you cannot establish a reliable hourly commitment, pay-as-you-go pricing avoids lock-in. Examples include one-time data processing jobs or short-term R&D projects.
• Workloads Planned for Immediate Modernization: If you are actively planning to re-architect and migrate a significant portion of your compute away from EC2 or Fargate (e.g., to containers on EKS or a serverless architecture using Lambda), a Savings Plan commitment could become stranded. It's better to complete the migration first and then commit based on the new architecture.
• Organizations with Cash Flow Constraints: The upfront payment options (especially All Upfront), while offering the highest discount, require a significant capital expenditure (CapEx). Companies that need to preserve cash and strictly operate on an operational expense (OpEx) model may find the pay-as-you-go model easier to manage, even if it's more expensive hourly.
• Legacy Environments Slated for Decommissioning: If you have a large, steady-state workload running on old instance types (e.g., M3, C3) that is scheduled to be fully shut down within the next 12 months, a 1-year commitment might not break even before the shutdown is complete.

In these cases, the flexibility and lack of commitment in On-Demand or Spot Instances outweigh the potential savings of a Savings Plan. The core principle is: if you cannot confidently predict your baseline compute usage for the next year, you should not commit to it.

Q6. What are the pricing models of AWS Reserved Instances and AWS Savings Plans, and how are discounts impacted by the type of payment made?

The pricing models for both RIs and Savings Plans are fundamentally about trading upfront financial commitment for a lower effective hourly rate. The golden rule is: the more upfront capital you provide AWS, the higher your effective discount will be.

AWS Reserved Instances (RIs) Pricing Models

RIs offer three payment options that directly impact your discount and cash flow:

a) All Upfront:

• Model: You pay the entire cost of the reservation term (1 or 3 years) in one single, upfront payment.
• Discount Impact: This option provides the highest effective discount (up to ~72% for a 3-year Standard RI). AWS rewards you most for providing all the capital immediately.

b) Partial Upfront:

• Model: You pay a portion of the total cost upfront and are billed a significantly discounted hourly rate for the instance for the duration of the term.
• Discount Impact: Offers a lower discount than All Upfront but a higher discount than No Upfront. It balances upfront cost with ongoing savings.

c) No Upfront:

• Model: You make no upfront payment and are simply billed a discounted hourly rate for the instance over the term.
• Discount Impact: Provides the lowest effective discount of the three options. It preserves cash flow but is the most expensive way to utilize an RI over the long term.

AWS Savings Plans Pricing Models

Savings Plans mirror the RI payment options but apply them to a dollar-hour commitment instead of a specific instance:

a) All Upfront:

• Model: You pay for your entire commitment (e.g., $10/hour for 3 years) in one payment.
• Discount Impact: Provides the highest effective discount for the commitment.

b) Partial Upfront:

• Model: You pay a portion of the commitment upfront and a discounted hourly rate for the remaining balance.
• Discount Impact: Offers a middle-ground discount, better than No Upfront but less than All Upfront.

c) No Upfront:

• Model: You make no upfront payment and are billed at the discounted Savings Plans hourly rate for your usage.
• Discount Impact: Provides the lowest effective discount but requires no initial capital. 

Part 3: Application & Flexibility of Savings Plans

Q7. Once a Savings Plan is purchased, how is it applied to your AWS setup?

AWS automatically applies Savings Plans to your eligible usage, with no manual assignment required. It works on an hourly basis in a specific priority order:

• Automatic Application: Each hour, AWS identifies your eligible compute usage (EC2, Fargate, Lambda) across your entire organization.
• Discount Application: It first applies any Reserved Instance discounts you have. Then, it applies your Savings Plan commitment to the remaining On-Demand usage that matches the plan's family and region.
• Priority by Discount: Usage is covered starting with the resources with the highest On-Demand rates first, ensuring you maximize your savings.
• Beyond the Commitment: Any usage that exceeds your Savings Plan commitment or doesn't match its terms is billed at the standard On-Demand rate.

You don't need to assign it to specific instances; AWS handles it seamlessly in the background.

Q8. Both AWS Reserved Instances and Savings Plans are commitments. If someone no longer needs that capacity after purchase, is there any way to get a refund?

No, you cannot get a cash refund. AWS does not offer refunds for canceled commitments. However, you do have options to minimize further loss, though they come with restrictions:

a) Reserved Instances (RIs):

Sell on the AWS Marketplace: If you have a Standard RI that you no longer need, you can try to sell it on the AWS Reserved Instance Marketplace to another customer. This is often difficult, and you will likely sell it at a loss.

• Modify: You can change the Availability Zone, scope (Regional vs. Zonal), or network platform of a Standard RI.
• Exchange: Convertible RIs can be exchanged for a different Convertible RI with a new attribute configuration (e.g., a different instance type or family). The new RI must have an equal or greater value, and the term resets.

b) Savings Plans:

• No Marketplace: There is no marketplace to sell Savings Plans.
• Limited Flexibility: If you change your mind within the first month—before the billing cycle—you can cancel the Savings Plan entirely. But the savings will also be lost, and your bill will revert to standard On-Demand rates.

The bottom line: Only commit to what you are very confident you will use. There is no easy "undo" button.

Part 4: Automation & Optimization of AWS Commitments

Q9. Is there a way for automation tools to handle AWS Reserved Instances?

Yes, absolutely. In fact, for any organization of significant size, using automation is essential to manage the complexity and risk of AWS RIs. These tools handle the entire lifecycle:

a) Recommendation: Analyze historical usage to recommend the optimal AWS RI type, instance family, term, and quantity to purchase to maximize coverage and savings.

b) Procurement: Automate the actual buying process based on those recommendations.

c) Management & Optimization: Continuously monitor your environment to:

• Ensure purchased RIs are being utilized.
• Identify opportunities to modify or exchange underutilized RIs.
• Track expiration dates and recommend renewals.

d) Waste Prevention: Alert you to AWS RIs that are going unused so you can take action (e.g., sell or exchange them) before the commitment is wasted.

Our tool CloudKeeper Auto is the only solution you need to manage AWS RIs and Savings. From auto-provisioning to shutting down idle instances to right-sizing, CloudKeeper Auto is a comprehensive AWS RI management tool . You get the convenience of on-demand flexibility with the pricing benefits of committed usage.

To Sum Up

For an organization that’s entering into cloud space with AWS, eventually, slightly scaling up from AWS to production workload scale, you would find yourselves opting for one or the other—Savings Plan or RI Plan. In 2025, AWS is actively pitching Savings Plans while AWS RIs are still available. Both have their utilities, but it’s essential to first have a clear understanding and visibility into your AWS infrastructure, then take an informed decision before entering into the appropriate contract.