Every team does the AWS cleanup sprint. Somebody deletes the obvious stuff, the bill dips, everyone feels good.
Three months later the bill is back, because the obvious stuff was never the problem. The waste that survives is the kind that doesn't look like waste: it's small, it's spread over regions nobody opens, and it has no owner.
Here are six kinds worth checking today. Each has a command you can run with read-only credentials. Run them per region (--region), because that's where most of it hides.
1. Unattached EBS volumes
You terminate an instance, and its data volume stays behind because "delete on termination" was off. You keep paying for the provisioned size whether or not anything reads it.
aws ec2 describe-volumes \ --filters Name=status,Values=available \ --query "Volumes[].[VolumeId,Size,CreateTime]" --output table Anything in available state and months old is a candidate. Snapshot it first if you're nervous.
2. Elastic IPs attached to nothing
Since February 2024, AWS charges for every public IPv4 address, in use or not, at $0.005 per hour. That's roughly $3.65 a month each. Small per IP, but orphaned ones pile up.
aws ec2 describe-addresses \ --query "Addresses[?AssociationId==null].[PublicIp,AllocationId]" --output table 3. Log groups that never expire
CloudWatch log groups default to never expire. A chatty Lambda from a project that ended two years ago is still storing every line.
aws logs describe-log-groups \ --query "logGroups[?retentionInDays==null].[logGroupName,storedBytes]" --output table Sort by storedBytes and set a retention period on the biggest ones. This is the safest cleanup on the list, because setting retention deletes nothing today.
4. Old snapshots
Snapshots are cheap one at a time. A nightly backup job that nobody ever pruned is 365 of them a year, per volume.
aws ec2 describe-snapshots --owner-ids self \ --query "Snapshots[?StartTime<='2025-01-01'].[SnapshotId,VolumeSize,StartTime]" --output table Change the date to suit you. Check that a snapshot isn't backing a registered AMI before you remove it.
5. Idle load balancers
A load balancer bills by the hour even with no traffic. Ones left over from a decommissioned service often have no healthy targets at all.
aws elbv2 describe-load-balancers \ --query "LoadBalancers[].[LoadBalancerName,LoadBalancerArn,CreatedTime]" --output table Then check each one's RequestCount metric in CloudWatch over the last 30 days. Zero requests for a month is a strong hint.
6. NAT gateways in quiet VPCs
NAT gateways charge an hourly fee plus data processing. A dev VPC that nobody has touched since the last offsite still pays the hourly fee.
aws ec2 describe-nat-gateways \ --filter Name=state,Values=available \ --query "NatGateways[].[NatGatewayId,VpcId,CreateTime]" --output table Cross-check against the VPC's traffic before you touch it.
Finding it is the easy half
Notice what these commands don't tell you: whose the resource is, what code created it, or what breaks if it goes. That's the part that makes people leave the volume where it is for another year.
A useful habit, before deleting anything:
- Find the owner. Check tags, then CloudTrail for who created it, then the repo for the IaC that declares it.
- Check dependencies. Is anything still referencing it?
- Make it reversible. Snapshot, stop instead of terminate, or set retention instead of deleting.
- Write down what you did, so recurrence is visible next quarter.
Doing this across 20 accounts
Running six commands in one account is a coffee break. Running them across every account and region, attributing each hit to an owner and to code, then keeping it from coming back, is a part-time job.
That's what I'm building Recoup for. It scans across accounts and regions with read-only credentials, shows the formula behind every cost estimate, attributes findings to owners and the code that created them, and generates tiered, reversible cleanup scripts. It never runs them for you. It also tracks whether waste you removed has come back.
It's in an invite-only pilot. To request a code, email hello@recoupify.in with the subject Recoup invite, and tell me how many AWS accounts you run. More at https://recoupify.in.
Your turn
Which of these six did you find in your own account? And what's the weirdest one that didn't make the list? I'll add the best ones to a follow-up.
