Loading...
Commit to capacity for one or three years and pay up to 70 percent less for it.
Quick Definition
Reserved Instances (RIs) are a commitment-based pricing model offering significant discounts in exchange for committing to specific instance usage over one or three years. Ideal for steady, predictable workloads, RIs reduce cost but require accurate forecasting to avoid underutilization.
Reserved Instances, or RIs, are a pricing agreement: you commit to a quantity of compute, traditionally a specific instance type in a specific region, for one or three years, and the provider discounts it 30 to 70 percent against on-demand rates. The hardware is identical; only the price and the promise change.
The discount is payment for predictability, and the risk is the same predictability: an RI you stop using keeps billing. The craft is matching commitments to genuinely stable usage, the floor of your demand curve, and choosing flexibility options wisely. Convertible RIs trade a few points of discount for the right to change instance families later; savings plans go further in the same direction.
Example. A team analyzes a year of usage and finds 60 instances running continuously. It reserves exactly that baseline for one year, keeps the variable layer on-demand, and cuts compute spend 38 percent with utilization above 97 percent.
Track utilization and effective savings rate monthly; unused commitments are how discounts become losses. The Cloud Cost Management Guide covers building a commitment strategy step by step.
Three-year terms discount more but bet on a distant future. Many teams ladder: three-year for the deepest stable core, one-year above it.
Convertible RIs can be exchanged, some marketplaces allow resale, and matching workloads can be moved onto the reservation. Prevention beats cure: commit below proven baseline.
Savings plans offer similar discounts with more flexibility and fit most compute. RIs persist for specific cases and some non-compute services.