An EC2 instance is not simply "on or off": it goes through several states during its lifetime, and each has important consequences, especially on your bill. Confusing "stop" with "terminate" is one of the mistakes that most often surprises beginners. Let's clarify this once and for all.

These are the states an instance goes through:

Let's look at the three states that really matter: running, stopped, and terminated.

When the instance is running, it is working and you pay for it: you pay for compute (per second or per hour depending on the type) and the associated storage.

This is the normal working state. Here is where your server serves your website, runs your application, etc.

Stopping (stop) an instance is like shutting down your home computer:

• The operating system shuts down.
• You stop paying for compute (the "processor rental").
• You keep paying for the disk (the EBS storage, which we'll see shortly), because your data is still stored.
• You can start it up again whenever you want and your data will still be there.

Real example: You have a test server that you only use during office hours. You stop it at night and on weekends. This way you stop paying for compute when you're not using it, but keep all your work. When Monday comes, you start it up and it's just as you left it. This can save more than 60% of compute costs.

An important detail when stopping: some things may change when you start it up again, such as the public IP (it may be different). To avoid this, use Elastic IPs, which we'll see in subchapter 4.4.

Terminating (terminate) an instance is to destroy it completely:

• The instance disappears.
• By default, its main disk is deleted and you lose all data that was on it.
• It cannot be undone. There is no recycle bin.

⚠️ Warning: "Stop" and "terminate" sound similar but are radically different. Stop = shut down (recoverable). Terminate = destroy (irreversible). Many newcomers terminate an instance thinking they were just shutting it down and lose their work. Read carefully before clicking.

When to terminate on purpose? When you no longer need the instance and want to stop paying for everything (compute and disk). This is the right thing to do to clean up test resources and avoid accumulating costs.

To really understand the lifecycle, you need to know where the data lives. An instance's disk is usually an EBS (Elastic Block Store) volume: a virtual hard drive attached to the instance.

The key is an option called "delete on termination":

• Enabled (default for the root disk): when the instance is terminated, the disk is deleted. Data lost.
• Disabled: the disk survives the termination of the instance and you can reuse or recover its data.

Best practice: For important data, don't rely solely on the instance's disk. Make snapshots (backups) of your EBS volumes regularly. A snapshot is a picture of the disk stored safely, from which you can restore. We'll see centralized backups in Chapter 26 (AWS Backup).

The big cost lesson: if you’re not using an instance temporarily, stop it (save compute). If you no longer need it, terminate it (save everything). Don’t leave unused running instances: it’s the number one cause of surprise bills.

• The key states are running (on, you pay for everything), stopped (off, you only pay for disk, data is safe), and terminated (destroyed, irreversible, root disk data deleted).
• Stop ≠ Terminate. Stop is recoverable; terminate is final. Read carefully before acting.
• EBS storage persists data; watch the "delete on termination" option and make snapshots of important things.
• To control costs: stop what you’re not using now, terminate what you no longer need.

In the next subchapter, we’ll see how to give your instances a fixed IP address (Elastic IP) and how to physically group them with Placement Groups.