EC2 Reserved instances - how do they work

EC2 Reserved Instances (RIs) work as a billing discount applied to the use of On-Demand Instances in your account, rather than as a physical allocation of instances. Here's how your scenario would play out:

1. You reserve 20x t3a.nano instances for a year.

2. For the first 6 months when you're using 40x t3a.nano:

• 20 instances would be billed at the discounted Reserved Instance rate.
• The additional 20 instances would be billed at the regular On-Demand rate.

3. For the next 6 months when you're using 0 instances:

• You would still be billed for the 20 Reserved Instances, even though you're not using them.

The reservation does not "average out" over the course of the year. It's applied on an hourly basis, up to the number of instances you've reserved. You pay for the entire term of a Reserved Instance, regardless of your actual usage. This means that in your scenario, you would not fully benefit from your RI purchase during the second half of the year when you're not using any instances.

Reserved Instances are best suited for steady-state usage. If your usage fluctuates significantly, as in your example, you might not fully optimize your cost savings. In such cases, it might be more beneficial to consider other pricing options like Savings Plans, which offer more flexibility, or to adjust your RI purchase to match your minimum consistent usage.
Sources
Reserved Instances
Reserved Instances for Amazon EC2 overview - Amazon Elastic Compute Cloud
Amazon EC2 Reserved Instances - Amazon EC2 Reserved Instances and Other AWS Reservation Models