← All articlesInfra and cloud

What we find in almost every AWS account

Transfers, bandwidth, sleeping instances, forgotten IPs, old backups. Nothing dramatic, but up to a fifth of the bill.

Julien Mauclair · ~3 min · intermediate
StackAWSEC2EBSFinOps

At Stralya, when we open an AWS account, we almost always find the same five leaks: data transfer costs between EC2 instances, bandwidth costs, instances running at 10% of their capacity, orphaned IP addresses and outdated backups. Added up, they can reach 20% of the bill.

01Why it is always the same ones

None of these leaks comes from one bad decision. They come from time passing. An instance sized for a peak that never came back. An IP address kept "just in case" after a server was deleted. Automatic backups nobody expires. A service moved to another Availability Zone without checking what the traffic between them costs.

They share one more thing: none of them breaks anything. The site runs, alerts stay quiet, and the line slowly grows on the bill. That is why you have to go looking for them, on purpose, with a list.

02The five leaks, and how to spot them

Transfer between EC2 instancesTraffic between Availability Zones, billed in each direction
BandwidthData served to the internet without caching or compression
Instances at 10%Oversized machines running almost idle
Orphaned IPsPublic IPv4 addresses still allocated but no longer used
Outdated backupsSnapshots kept long after they stopped being useful

Cross-zone transfer is the quietest leak. AWS charges for data moving from one Availability Zone to another in the same Region, both in and out, as the EC2 pricing page explains. An application and its database in two different zones pay for every request twice. My article on NAT gateway cost shows a related case.

Orphaned IP addresses cost little each, but they add up: since 2024, every public IPv4 address is billed $0.005 per hour, in use or idle, according to the VPC pricing page.

03The commands for a first pass

These AWS CLI commands do the first sort. They delete nothing: they list.

leak-audit.shBASHCopy
# Elastic IPs allocated but associated with nothing
aws ec2 describe-addresses \
--query 'Addresses[?AssociationId==null].[PublicIp,AllocationId]' \
--output table
# Snapshots created before a given date
aws ec2 describe-snapshots --owner-ids self \
--query "Snapshots[?StartTime<='2026-05-01'].[SnapshotId,VolumeSize,StartTime]" \
--output table
# Average CPU of one instance over 14 days, per day
aws cloudwatch get-metric-statistics \
--namespace AWS/EC2 --metric-name CPUUtilization \
--dimensions Name=InstanceId,Value=i-0123456789abcdef0 \
--start-time 2026-10-26T00:00:00Z --end-time 2026-11-09T00:00:00Z \
--period 86400 --statistics Average

For transfer and bandwidth, the easiest tool is Cost Explorer, grouped by usage type: lines containing DataTransfer-Regional-Bytes are cross-zone traffic, lines with DataTransfer-Out-Bytes are data sent to the internet. Filters and grouping are described in the Cost Explorer documentation.

04Where to start

In order, from easiest to most delicate:

·orphaned IPs: check that no DNS record points to them, then release them;
·outdated backups: set a retention period, then delete what is older;
·instances at 10%: drop one size, watch for a week, repeat if needed;
·bandwidth: put a cache in front of what is served often;
·cross-zone transfer: bring chatty services closer together, keeping high availability where it is really needed.

The first two take about an hour and carry almost no risk once the checks are done. The next three require understanding the application, and that is where most of the savings are. My guide to cutting an AWS bill covers the full method, and the article on common leaks gives more leads.

05FAQ

How much can you recover on an AWS bill?

On the accounts Stralya audits, these five leaks add up to as much as 20% of the bill. The figure depends a lot on the age of the account and on how many people have created resources in it.

Is it risky to delete snapshots or IPs?

Yes, if you do it blindly. For an IP, first check that no DNS record or allowlist uses it. For a snapshot, check that it is not the only copy of a volume still in use.

How often should you repeat this review?

Once a quarter is enough for most accounts. Leaks come back as soon as new projects get created, so a regular pass costs less than a big yearly clean-up.

In short

AWS account leaks are rarely dramatic: they are small oversights that pile up. Looking for them once a quarter is often enough to win back a visible share of the bill.

Talk about your project ↗
Julien MauclairJulien Mauclair is the co-founder and CTO of Klickbee, where he builds the agency’s websites, tracking and AI automations, and develops Klickbee Tracking. He also runs Stralya, which designs and runs AWS infrastructure for scale-ups. Before that, he built the entire AWS infrastructure of an e-commerce business, since sold.

Read next