Reserved Instances

A reserved instance is a commitment to pay for a given amount of cloud compute over a fixed term — typically one or three years — in exchange for a lower rate than on-demand pricing. Related products go by other names (savings plans, committed-use discounts) and differ in how much flexibility they preserve, but the shape is the same: you trade optionality for a discount, and the discount grows with the length of the term and with how specific the commitment is. Reservations sit at one end of a spectrum. On-demand costs the most and can be turned off at any moment. Spot capacity costs the least and can be taken away from you at short notice. Reserved capacity is cheap and always available, but you pay whether or not you use it, which turns an operating cost into something closer to a fixed one. The practical rule follows directly: reserve the floor, not the peak. Commit to the baseline your workload never drops below, cover the variable part on-demand, and put interruptible batch work on spot. Two failure modes are common. Over-committing during a growth forecast that does not arrive leaves you paying for idle capacity for the rest of the term. And committing to a narrowly specified instance family can outlive the architecture that needed it — a migration to a different processor type or to managed services can strand a reservation, so the commitment should be reviewed alongside any planned platform change rather than treated as pure finance.