Short answer: partly
Does a stopped EC2 instance still cost money?
Stopping is not deleting, and AWS bills the difference. Here is exactly which charges stop and which carry on.
Stopping an EC2 instance stops the instance-hour charge immediately. It does not stop anything attached to it. You keep paying for the root and data EBS volumes at $0.08 to $0.10 per GB-month, for any snapshots at $0.05 per GB-month, and for any associated Elastic IP or public IPv4 address at $3.60 a month. A stopped t3.large with a 100 GB gp2 root volume costs about $10 a month to leave stopped and $0 to terminate.
What stops and what does not
| Charge | While running | While stopped |
|---|---|---|
| EC2 instance hours | Billed | Not billed |
| EBS root and data volumes | Billed | Billed in full |
| EBS snapshots | Billed | Billed |
| Elastic IP / public IPv4 | $3.60 per address per month | $3.60 per address per month |
| Data transfer | Billed on use | None, nothing is moving |
| Provisioned IOPS on io1/io2 | Billed | Billed |
| Licence-included Windows / RHEL | Billed | Not billed |
On-demand list prices, 730 hours a month. Instance-store volumes are a special case: their data is destroyed on stop, so there is nothing left to bill.
Worked example
A t3.large in eu-west-2 runs at roughly $0.0928 an hour, which is about $68 a month. Attached: a 100 GB gp2 root volume at $0.10 per GB-month, so $10, and one Elastic IP at $3.60.
Running: about $81.60 a month. Stopped: $13.60 a month, because the volume and the address do not care whether the instance is running. Terminated with the volume and address released: nothing.
The trap is that stopping feels like it solved the problem. Fifty stopped instances from old projects, each with a root volume, is $500 a month of nothing.
The useful cases for stopping rather than terminating
Scheduled shutdown of non-production. Stopping dev, staging and QA outside working hours removes roughly 70% of their instance cost and keeps the machines exactly as they were. This is the single safest cost change available on AWS and it needs no architecture work.
Anything you might genuinely need back within weeks. Below about a month, the EBS charge is cheaper than the labour of rebuilding.
Beyond that, take a snapshot and terminate. A 100 GB snapshot costs $5 a month against $10 for the live volume, and snapshots are incremental, so the real figure is usually much lower.
Put numbers on it
Frequently asked questions
Does a stopped instance still count against my Savings Plan?
No. Savings Plans apply to compute usage, and a stopped instance generates none. If a large part of your fleet stops overnight, size the commitment against the hours you actually run, not the peak.
How do I stop being charged completely?
Terminate the instance, and check that the "delete on termination" flag was set on every attached volume, because it defaults to true for the root volume and false for data volumes. Then release the Elastic IP. An unassociated Elastic IP still bills at $3.60 a month, which is the most common orphan we find.
What about hibernation?
Hibernated instances write RAM to the root volume, so the volume gets larger and the storage charge goes up slightly. Instance hours still stop. It is a faster restart, not a cheaper one.
Do stopped RDS instances work the same way?
Similar but worse. RDS stops the instance charge but keeps billing storage and backups, and AWS automatically restarts a stopped RDS instance after 7 days. If you are stopping RDS to save money, you need a scheduled job to stop it again every week, or you need to delete it with a final snapshot.
Keep reading
Prices checked against AWS list rates on 2 August 2026. AWS changes prices; treat every figure here as a close approximation rather than a quote.
£499 fixed. Free scan first. 20%+ found or it’s free.
The scan is read-only — a role you create and delete, no keys shared — and shows your estimated monthly saving before anyone talks about money.