Technical guide
How EC2 pricing actually works
A practical breakdown of EC2 instance-hours, purchase options, attached resources, and the assumptions behind a useful estimate.
Updated 2026-09-05 · Data verified2026-09-05
The direct calculation
For an On-Demand Linux instance, the compute subtotal is the hourly instance rate multiplied by running instance-hours. One t3.medium running for 730 hours at $0.0416/hour is $30.368 before storage, transfer, public IPv4, load balancing, monitoring, support, taxes, or discounts.
What changes the bill
Region and operating system can change the rate. Stopping an EBS-backed instance stops compute billing but normally leaves attached EBS volumes billable. Reserved Instances and Savings Plans change the effective compute price and introduce term, payment, and coverage considerations; they do not make every attached service free.
Estimation checklist
Model instance count and hours first. Then add EBS capacity and provisioned performance, outbound transfer, public IPv4 time, load balancers, and managed-service dependencies. Compare the result with billing exports before using it as a forecast.