You're Paying for Reserved Instances You're Not Using — And Probably Don't Know It
Reserved Instances are one of the best deals in cloud computing. Commit to one or three years of usage for a specific instance type and region, and AWS offers discounts of up to 72% compared to on-demand rates. Savings Plans extend the model with more flexibility — commit to a dollar amount of compute spend per hour, and the discount applies across a wider range of services and instance types.
The problem isn't the discount. The problem is what happens when the workload changes — and the commitment doesn't.
The production database you bought an RDS Reserved Instance for got migrated to Aurora Serverless. The EC2 instances your Reserved Instances were covering got replaced by Lambda functions. The application you committed capacity for was deprecated in Q2. The Reserved Instances are still running. The bill is still arriving. And the capacity you're paying for is no longer doing anything useful.
The Scale of the Problem
The numbers on unused cloud commitments are striking.
According to ProsperOps' 2024 Effective Savings Rate Benchmarks report, 53% of organizations analyzed were not utilizing any commitment instruments — no Savings Plans, no Reserved Instances — for compute. The median organization had an Effective Savings Rate of 0% or lower, meaning they were capturing no benefit from the commitment discounts available to them.
Among organizations that do use commitments, underutilization is the persistent challenge. The 2024 State of FinOps report from the FinOps Foundation found that managing commitment-based discounts rose to the second highest priority for FinOps practitioners, up sharply from prior years — a signal of how widespread the problem is and how urgently teams feel the need to address it.
Meanwhile, Flexera's annual report consistently finds that organizations estimate they waste approximately 28–32% of their total cloud spend, with unutilized or underutilized commitments a significant contributor.
The pattern is almost universal: organizations buy commitments with optimism about their workloads, workloads change in ways nobody fully anticipated, and the commitments outlast the workloads they were purchased to cover.
Why Underutilization Is So Easy to Miss
Reserved Instances and Savings Plans don't fail visibly. They don't generate an alert. They don't appear in your cost anomaly detection as a spike. They don't trigger a billing alarm.
They just quietly pay, month after month, for capacity you may no longer be using. The cost is consistent and predictable — which is part of what makes it invisible. Anomaly detection looks for unusual changes in spending patterns. A Reserved Instance that has been underutilized for six months doesn't look anomalous; it looks like a stable fixed cost.
The AWS tools for tracking utilization — Cost Explorer's Reservation Utilization and Savings Plan Utilization dashboards — exist, but they require deliberate attention. They're not surfaced by default in billing alerts. They don't push notifications when utilization drops below a threshold. They wait for you to go look.
There's also a psychological dynamic at play. Reserved Instances are usually purchased with careful analysis. They represent a deliberate, approved decision — often reviewed by a finance team. Once purchased, there's organizational inertia against treating them as a problem. Acknowledging that a Reserved Instance you bought eighteen months ago is now underutilized requires acknowledging that the workload assumptions behind that purchase were wrong. That's a harder conversation to initiate than flagging an unexpected on-demand charge.
The Three Failure Modes of Commitment Management
Most underutilization traces back to one of three structural problems.
Purchase decisions that outpace workload stability. Commitments purchased for workloads during periods of rapid architectural change — a microservices migration, a database consolidation, a platform modernization — frequently outlast the specific configurations they were designed to cover. The commit was reasonable at the time; the workload evolved faster than the three-year term anticipated. This is particularly common with Standard Reserved Instances, which are tied to a specific instance family, region, operating system, and tenancy, and can't be modified if the workload moves.
No coverage tracking after purchase. Many organizations treat commitment management as a periodic exercise rather than an ongoing practice. They review RIs and Savings Plans at renewal time — and not in between. During the gap, utilization can deteriorate significantly without triggering any review. By the time the commitment comes up for renewal, it's been underutilized for a year.
Overcommitment to avoid on-demand exposure. The instinct to maximize committed spend coverage — to minimize the fraction of workload running at on-demand rates — can lead to purchasing more commitment than actual usage justifies. Organizations that aggressively cover 100% of their compute with commitments are almost certainly overcommitting in at least some service areas, particularly for workloads with variable or unpredictable utilization.
The Metrics That Actually Matter
Coverage and utilization are the traditional metrics for commitment management. They're useful but incomplete.
Coverage measures what percentage of your eligible spend is covered by a commitment. High coverage means you're capturing discounts across most of your compute. But high coverage doesn't tell you whether those commitments are being used efficiently — an over-purchased commitment with 60% utilization has high coverage but significant waste.
Utilization measures how much of a purchased commitment is actually being used. A commitment at 100% utilization is generating maximum value. A commitment at 40% utilization is delivering 40% of the savings it was purchased to generate, while the remaining 60% is committed spend covering capacity that isn't being consumed.
Effective Savings Rate (ESR) — a metric formalized by the FinOps Foundation and ProsperOps — combines both dimensions to produce a single measure of what you're actually saving relative to what you would pay on-demand across your full compute footprint. ESR is harder to game than coverage or utilization in isolation, and it surfaces the true financial outcome of your commitment portfolio: not whether you have commitments, but whether those commitments are generating real savings.
What a Healthy Commitment Practice Looks Like
Organizations that manage commitments well share a few practices that distinguish them from those that don't.
They review commitment utilization on a regular cadence — monthly at minimum — rather than only at renewal time. They track utilization at the service level (EC2, RDS, ElastiCache, Savings Plans separately) rather than only at the aggregate, because underutilization in one service can be masked by strong utilization in another.
They use Convertible Reserved Instances where possible instead of Standard RIs. Convertible RIs offer up to 66% discount rather than 72%, but they can be exchanged for RIs with different attributes during the term. For workloads with any chance of changing — different instance type, different region, different OS — the exchange flexibility is worth the small discount reduction.
They set up utilization alerts at meaningful thresholds — 85%, 70%, 60% — so that deteriorating utilization triggers a review before it becomes a waste problem rather than after.
And they treat commitment decisions as ongoing portfolio management rather than one-time purchases. The question isn't only "should we buy this commitment?" It's also "what's in our commitment portfolio now, and is it still appropriate for our current workloads?"
Reserved Instances done well are among the most reliable cost savings available in cloud. Reserved Instances done poorly are a reliable source of waste that compounds quietly for years. The difference is whether someone is watching the utilization — and whether the tooling makes that watching easy rather than requiring deliberate effort.
Reduce tracks Reserved Instance and Savings Plan utilization and coverage across your entire AWS commitment portfolio — with alerts when utilization drops and visibility into exactly where the waste is landing.