2025 Rate Optimization Insights: AWS Compute

TL;DR

Rate optimization insights and Effective Savings Rate (ESR) benchmarking data can help you achieve better savings outcomes for AWS compute.

• 60% or more of the overall AWS usage in the survey pool was compute, with ~90% coming from EC2 versus Lambda and Fargate.
• The median AWS Compute ESR rose from 0% in 2023 to 15% in 2024, due to more organizations leveraging commitments. 64% used RIs/SPs in 2024, an increase from 45% in 2023.
• Median coverage also jumped to 55% in 2024, as opposed to only 28% in 2023.
• However, FinOps practitioners assumed greater commitment risk to realize those savings. For example, organizations at the top percentiles hit a coverage ceiling as it became harder to improve ESR while accounting for commitment risk.
• In 2024, 51% of organizations using commitments utilized infrequent batch purchases of AWS Savings Plans over sophisticated strategies and/or diversified commitment portfolios.
  • The most popular type of commitment was the 3-year Compute Savings Plan, which 50% of organizations used in 2024, because it is easy to implement, covers a broad range of compute services, and requires no ongoing management.
• FinOps maturity, using ESR as a proxy, increased with compute usage; in other words, compute usage and ESR were correlated:
  • The smallest segment of organizations (i.e., those with less than $500K in annual compute usage) had the biggest opportunity to improve ESR, as they performed no better than paying on-demand with a 0% median ESR.
  • The largest segment of organizations (i.e., those with $10M or more in annual compute usage) achieved 38% median ESR in 2024, the highest vs. other segments, but saw minimal year-over-year gains due to commitment risk.
• To maximize ESR and minimize risk, commitment flexibility is crucial; however, many FinOps practitioners are unfamiliar with the strategies, processes, and/or technologies required to deliver world-class rate optimization outcomes.
• To consistently sustain a high ESR with commitment flexibility in a dynamic environment, automation is critical. Third-party FinOps platforms, such as ProsperOps, leverage algorithms/AI/automation to optimize both ESR and commitment risk.

Introduction

As of this report, AWS is still the largest cloud provider, with an annual estimated revenue of $117B based on its Q1 2025 financials. ProsperOps analyzed March 2025 customer data and found that 53% of total AWS spend (defined as the actual cost paid after discounts have been applied) was attributed to Amazon EC2, AWS Lambda, and AWS Fargate. This suggests that 60% or more of usage (defined as the cost before discounts have been applied) was compute. Therefore, optimizing AWS compute spend delivers an outsized impact on reducing your overall cloud cost.

When it comes to the specific compute service consumed, EC2 remains the standard. EC2 accounted for nearly 90% of compute usage, though many organizations also utilized serverless compute. This ratio was consistent from 2023 to 2024. Serverless may become more prevalent in the future, but its adoption is dependent on how organizations architect their environments, the nature of their workloads, and other business needs.

To reduce cloud costs without impacting engineering, FinOps practitioners can leverage a variety of tools, such as commitments, volume-based discounts, Enterprise Discount Programs (EDP)/Private Pricing Agreements (PPA), and/or alternative pricing models (e.g., Spot Instances). This activity is known as rate optimization.

Optimizing rates using commitments, such as Savings Plans (SPs) and Reserved Instances (RIs), reduces costs in exchange for committing to specific resources and/or spend amounts as part of a contract term. This is an ongoing process because computing environments tend to be dynamic and require continual data analysis from AWS Cost and Usage Reports and AWS Cost Explorer to understand patterns. Rate optimization with commitments becomes increasingly complex as the magnitude and volatility of the computing environment increase.

Key Statistics for 2024

Methodology

This report explores factors influencing rate optimization outcomes (e.g., compute services used, commitment coverage, etc.), revealing trends in ESR, RI/SP consumption, and commitment risk management. Our intention with these findings and insights is to help FinOps practitioners improve the effectiveness of their work.

We examined approximately $3 billion in AWS compute usage across Amazon EC2, AWS Lambda, and AWS Fargate, using AWS Cost and Usage Reports (CUR) and AWS Cost Explorer data. To ensure statistical reliability and reveal meaningful trends, we calculated ESR performance using a 24-month lookback period. This approach increases sample size, reduces short-term noise, and highlights directional shifts in cloud savings behavior over time. “2024” represents monthly lookback data spanning from January 2023 through December 2024, and “2023” represents data spanning from January 2022 through December 2023.

Effective Savings Rate

Traditional “input” metrics used to assess rate optimization effectiveness (i.e., coverage, utilization, and discount rate) are useful, but fail to provide an accurate measure of your net return on investment when they are evaluated in isolation. Effective Savings Rate (ESR), on the other hand, is an “output” FinOps metric adopted by the FinOps Foundation, AWS, practitioners, and vendors to accurately measure the success of rate optimization activities. ProsperOps’ rate optimization solution, Autonomous Discount Management, is designed to maximize ESR, and this metric is tracked in our console dashboard.

ESR can be scoped to measure the impact of any type of discount, including privately negotiated discounts (EDP/PPA) and Spot Instances. In this report, ESR is calculated using the total savings generated from Savings Plans and/or Reserved Instances off the on-demand rate for AWS compute services (EC2, Lambda, and Fargate). This is consistent with how AWS measures and reports on RI and SP savings. Limiting the scope of ESR to just SPs/RIs allows for equal comparison across AWS organizations, independent of their size and spend. EDP/PPA discounts are additive and, if included in the scope, would further increase savings and ESR.

---

Observations

Here are several observations from our analysis on AWS compute usage.

Compute ESR performance increased

The median (50th percentile) ESR rose from 0% in 2023 to 15% in 2024, due to broader adoption of commitments and coverage increasing.

However, improvement year-over-year was uneven across populations. The bottom 25th percentile remained at 0%, indicating no or ineffective rate optimization. At the top, progress was marginal; ESR at the 98th percentile increased just one percentage point, from 46% to 47%. This suggests that at high ESR levels, incremental gains are harder to achieve with traditional optimization strategies and processes. See last year’s report.

More AWS organizations leveraged commitments

Related to the previous point, more AWS organizations were using commitments overall. In 2023, 45% of AWS organizations surveyed were utilizing commitments to some degree. In 2024, that figure increased to 64%.

Median ESR pulled closer to the average

As more organizations used commitments to realize savings, the gap between the median and average ESRs narrowed. In 2023, half of organizations surveyed had ESRs of 0% or lower, dragging the median down, while top performers raised the average. By 2024, more organizations had achieved ESRs greater than 0%, bringing the median closer to the average.

Commitment coverage increased

Another reason ESR performance improved was that commitment coverage increased year-over-year. Average coverage (i.e., the percentage of compute usage covered by commitments) grew from 37% in 2023 to 47% in 2024, while the median significantly increased from 28% to 55% over the same period.

Note: The median exceeded the average in 2024 due to more organizations covering above 0%, whereas the average was suppressed by those remaining organizations with no coverage at all.

All things being equal, higher coverage is beneficial. However, high levels of commitment coverage can become problematic when usage patterns are volatile. We discuss the risks and mitigation steps in the following section.

Savings Plan adoption grew

While the absolute number of both SPs and RIs increased year-over-year, organizations allocated a greater share of their commitment portfolios to Savings Plans in 2024. AWS encourages organizations to leverage Savings Plans since they are easier to implement, cover/adjust to a broad range of compute services, and require no ongoing management once purchased.

Compute Savings Plans were the most popular commitment

The most popular type of Savings Plan used was the Compute Savings Plan (CSP). The CSP can float across EC2, Lambda, and Fargate automatically, regardless of instance family, size, region, operating system, and tenancy. This enables engineering teams to have more autonomy and fewer architectural limitations. Organizations were also comfortable with committing to three-year terms, which offer greater discounts in exchange for longer-term commitments—and potentially higher risk.

Standard Reserved Instances (SRIs) were the next most popular commitment, followed by Convertible Reserved Instances (CRIs). EC2 Instance Savings Plans (EC2 ISPs) were the least popular because they cannot float across regions, are restricted to the EC2 instance family, and have no marketplace option to address the commitment risk.

Organizations defaulted to “easy” rate optimization strategies

In 2024, 51% of AWS organizations taking advantage of commitments were only using Savings Plans, compared to 34% using both Reserved Instances and Savings Plans, and 15% using only Reserved Instances. A Savings Plan-only strategy, that is purchased infrequently in batches, is the simplest rate optimization strategy to execute, although it does not generally drive a high ESR outcome.

---

ProsperOps Perspectives

Commitment flexibility is key to achieving top-percentile ESR outcomes

As the data shows, 51% of organizations used a Savings Plan-only strategy. While Savings Plans are extremely effective tools for rate optimization, the common implementation strategy is infrequent batch purchases, which can limit ESR and carry significant commitment risk in volatile environments. This is because there is no native way to adapt the SP post-purchase when usage changes, apart from the limited return policy. Organizations intuitively understand that commitments carry risk and covered only 47% of their usage with commitments in 2024.

To mitigate commitment risk, your rate optimization strategy should factor in the concept of flexibility. Flexibility is the ability to adapt commitments to match dynamic usage. There are two dimensions of flexibility: horizontal and vertical. Horizontal flexibility, which is technical in nature, refers to the commitments’ ability to shift across different services or resource types based on attributes, such as product, instance family, and region. This is something Compute Savings Plans excel at and why they are so popular. Vertical flexibility, which is financial in nature, describes the ability to increase or decrease the total value of commitment to match increasing or decreasing usage. Having this type of flexibility mitigates the risk of under- and over-commitment from future usage changes. However, vertical flexibility is limited with infrequent batch-purchase strategies.

Organizations may undervalue vertical versus horizontal flexibility, as the latter is very apparent to engineering teams (i.e., “my commitment will apply, no matter what resource I use”). However, vertical flexibility is just as important, since one of the biggest ongoing challenges facing FinOps practitioners is managing commitments when usage levels fluctuate, whether predicted or not.

Achieving maximum flexibility requires automation

Unfortunately, many organizations have limited to no rate optimization automation, relying instead on infrequent and immutable batch commitment purchases executed manually. Because there is little to no vertical flexibility with this strategy, those organizations tend to commit too conservatively, thus paying more with on-demand rates, resulting in suboptimal ESR outcomes.

We believe organizations need automation to apply advanced techniques that embed flexibility into their rate optimization strategy. There are several techniques which can be automated to unlock flexibility in every commitment instrument. For example, Adaptive Laddering with Compute Savings Plans can deliver both native horizontal flexibility and vertical flexibility through the purchase of commitment “rung” increments.

Automation enables this vertical flexibility in multiple ways:

• High-frequency purchases: The more frequent the rung purchases, the more opportunities to adapt the ladder to changing usage patterns. For example, making purchases hourly creates multiple daily adjustment points, which increase ESR when usage is on the rise.
• Dynamic rung size adjustment: In most modern cloud compute environments, usage is variable. With frequent purchases, maximizing coverage and utilization requires ongoing effort. This includes continuously evaluating historical usage, predicting future usage, and adjusting the rung size (up or down), or allowing commitments to expire completely.
• Optimal coverage for cyclical workloads: For workloads with recurring dynamic usage patterns, overall commitment should target levels that purposefully forgo perfect utilization because doing so maximizes ESR (i.e., not just covering up to the usage trough). Continually recalculating this optimal commitment point with high-frequency rungs maximizes ESR.
• Ladder composition and balance: Many times, having balanced and symmetrical rungs of commitment makes the most sense. However, during periods in which usage changes predictably (e.g., weekends, Black Friday, etc.), having asymmetrical commitment rung sizes that match these drops allows coverage to track usage more closely. Managing high-frequency ladders with the ideal rung composition and balance increases ESR.

Other strategies leverage diversified portfolios of Savings Plans and Reserved Instances with automation to unlock additional vertical and horizontal flexibility. Depending on the environment, organizations can achieve top-percentile ESR outcomes by leveraging the right mix of strategies.

---

Considerations for FinOps Practitioners

Every organization can benefit from better rate optimization with the right strategy and capability

The magnitude of compute usage tends to be correlated with FinOps maturity. In the scatterplot below, we grouped organizations into three segments based on similar levels of compute usage.

Organizations with higher compute usage (x-axis) generally had higher ESRs (y-axis) and less variability in the ESR outcome. In contrast, smaller-spending organizations likely did not have the FinOps resources, processes, or incentives in place and generated low/negative ESRs.

While there were material differences in ESR outcomes across the different segments, as shown in the graph below, one common theme is that there is room for improvement for every organization, independent of FinOps maturity.

If you know your ESR, which segment do you fall into? Read on to learn how you can increase your ESR.

Smallest Usage Segment

Organizations in the smallest segment (red line, above), with less than $500K of annualized compute usage, had the lowest median ESR (0%) of any group. This segment may not be using commitments at all or utilizing them so poorly that any discount benefit is zeroed out.

Note: Some companies in this segment generated negative ESRs, which means that they would have been better off paying on-demand rates than using commitments.

This segment also had the highest standard deviation in ESR, which was likely the result of:

• Cloud optimization not being a top priority, given the modest level of spend
• Fewer internal resources, expertise, and time to dedicate to optimization
• Uncertainty about the long-term need for cloud, leading to more on-demand costs

Advice: Start with a simple, conservative rate optimization strategy to achieve an immediate cost reduction. For example, a one-year commitment with a 25% discount has a 9-month break-even point versus paying the on-demand rate.

Middle Usage Segment

Organizations in the middle segment (green line, above), with annualized compute usage between $500K and $10M, achieved decent ESR outcomes, but have significant room to improve both savings and commitment flexibility.

The median AWS Compute ESR in the middle segment rose to 23% in 2024, up from 20% in 2023. This increase was likely due to organizations focusing on operational efficiency and the standard tools and processes for rate optimization.

Advice: Aim to double your ESR (40%-46%). But to jump from a beginner/intermediate strategy to a more advanced strategy, you’ll need to build commitment flexibility into your program early.

Largest Usage Segment

For organizations in the largest segment (yellow line, above), with $10 million or more annualized AWS compute usage, cloud cost is already a top P&L line item and a strategic priority. These organizations have most likely invested in FinOps capabilities to a greater degree than the other segments and are achieving the ESR results to show for it. ESR was the highest for this segment, but increased just one percentage point, from 37% in 2023 to 38% in 2024.

However, incremental improvements are challenging as these organizations are already operating with very high levels of commitment coverage and associated risk. For example, the median commitment coverage for this segment in 2024 was 90% of usage. If and when cloud consumption growth stalls, a drop in usage can create an overcommitted state and lower their ESRs.

This is why commitment flexibility is critical. Even a modest ESR improvement can drive material incremental savings. For example, a five-percentage-point ESR gain on $10M of annual compute usage results in $500K of incremental savings annually, which can be reinvested into new projects, FTEs, or margins.

Advice: Prioritize the automation of rate optimization to add commitment flexibility above and beyond what you have in place. That flexibility allows you to safely cover more of your usage and increase ESR to drive incremental savings.

In summary, most organizations are still in the early innings of implementing their FinOps program, and there is still more optimization work to do than capacity to do it. Simplistic strategies, executed manually, cannot deliver consistently high ESR outcomes in dynamic environments. The key to top-percentile ESR outcomes in 2025 is commitment flexibility, and the key to commitment flexibility is automation.

If you’d like to learn more about ProsperOps and how we implement rate optimization strategies to accelerate your FinOps program, request a demo.

---

About ProsperOps

ProsperOps is the leading FinOps automation platform for cloud cost optimization on Amazon Web Services (AWS), Google Cloud, and Microsoft Azure. Eliminating waste and achieving cost-saving goals are challenging when cloud usage is elastic, but commitments are inelastic. Founded in 2018, ProsperOps reduces costs by synchronizing rate optimization with workload optimization, eliminating waste and boosting cross-team efficiency for FinOps. Our platform drives world-class Effective Savings Rates and mitigates Commitment Lock-In Risk for our customers.