EC2 Pricing Models Explained: On-Demand, Savings Plans, Reserved, Spot
5 min read · Updated 2026-10-25
EC2 offers four purchasing models and the exam tests the match between workload shape and pricing model. Stability and interruptibility are the deciding factors.
The options
- On-Demand — full price, no commitment; use for unpredictable or short-term workloads.
- Savings Plans — commit to $/hour of usage for 1–3 years; applies across EC2, Fargate, Lambda. Flexible and broadly applicable.
- Reserved Instances — commit to a specific instance configuration for 1–3 years; larger discount, less flexibility (Standard vs Convertible).
- Spot — up to ~90% off spare capacity; can be interrupted with a 2-minute warning; for fault-tolerant work only.
- Dedicated Hosts — physical server for licensing/compliance; expensive, not a discount play.
Decision logic
Steady 24/7 production → Savings Plans or RIs. Unpredictable or dev/test → On-Demand. Interruptible batch/CI/stateless → Spot. The classic trap: buying Spot for a database that cannot tolerate interruption, or leaving On-Demand on a stable workload.
Right-sizing comes first: discounting an oversized instance locks in waste. Use Compute Optimizer to fit the instance to the workload before committing.