.png)
How It Works
A Reserved Instance (RI) is a billing contract with a cloud provider. You commit to paying for a set amount of compute, database, or other capacity in exchange for a significant discount versus on-demand rates. AWS offers up to 72% off with Reserved Instances, Azure Reservations offer up to 72% off, and GCP Committed Use Discounts offer up to 57% off. The discount only delivers value when the commitment is actually consumed by a running workload. When a reserved resource sits idle, or when the instance type or region it covers no longer matches your running workloads, the provider still charges you the reserved rate. You pay for capacity your infrastructure never uses.
Why It Matters for Cloud Cost
Unused Reserved Instances are one of the most direct and preventable forms of cloud waste. Teams typically discover the problem after the fact: a workload was decommissioned, migrated, or downsized without anyone adjusting the underlying commitment. Because commitments run on fixed terms, the charge continues until the term expires. The financial impact compounds quietly. A single unused RI for a large database instance can cost tens of thousands of dollars over a 12-month term with zero savings to show for it. Across a multi-account organization with dozens or hundreds of commitments, unmonitored utilization gaps can consume a meaningful portion of the expected savings. Poor commitment hygiene also distorts reporting: gross savings figures look worse, finance loses confidence in cloud cost projections, and the case for further optimization becomes harder to make.
Key Characteristics
- Unused Reserved Instances generate no discount benefit because no workload is consuming the committed capacity.
- The financial loss is fixed and time-bound: the provider charges the reserved rate for the full remaining term regardless of usage.
- Underutilization can originate from workload decommissioning, rightsizing, migration, or over-purchasing at initial commitment time.
- On AWS, Instance Flexibility (Convertible RIs and Compute Savings Plans) can reduce stranded capacity risk by allowing some reallocation across instance families.
How Usage AI Handles This
Usage AI monitors commitment utilization daily and adjusts coverage through Autopilot, so commitments track your actual workloads rather than a static snapshot from months ago. For any underutilization that does occur on commitments purchased through Usage AI, the Guaranteed Buyback ensures you receive cashback plus credits rather than absorbing the loss.
See how Usage AI saves 30 to 50% on AWS, GCP, and Azure.
Common Questions
1. What is the difference between an unused Reserved Instance and an underutilized one?
An unused RI has zero consumption against the commitment for a given period. An underutilized RI is being partially consumed but not fully, so you capture some discount benefit but still pay for unused capacity. Both erode your effective savings rate, but unused RIs represent the highest-severity waste because the discount delivers nothing.
2. Can you sell or transfer an unused Reserved Instance?
AWS operates an RI Marketplace where Standard Reserved Instances can be listed for resale before their term ends. Not all instance types are eligible, Convertible RIs cannot be sold on the marketplace, and the resale process requires manual effort with no guarantee of a buyer. Azure and GCP do not have equivalent secondary markets for their commitment products.
3. How do Savings Plans compare to Reserved Instances for avoiding unused commitments?
AWS Compute Savings Plans apply a discount rate to any eligible compute spend across EC2, Fargate, and Lambda, rather than locking to a specific instance type or region. This flexibility makes it harder to end up with stranded coverage when workloads change. Usage AI uses both Savings Plans and Reserved Instances depending on the service and coverage need, optimizing the mix to minimize utilization gaps.