Underused EC2 instances cost

How to detect underused EC2 instances

The best signal is a running EC2 instance whose CPU stays consistently low for long enough to suggest steady overprovisioning rather than a brief quiet period.

Cloud Waste Hunter flags running EC2 instances whose average CPUUtilization stays at or below 5% across the last 14 days, or across the shorter period since launch. The implementation also requires at least three metric datapoints so a brief metrics gap does not create noise.

That means this page is about sustained low CPU, not full workload profiling. The detector is strongest when you want to find obvious overprovisioning and forgotten compute that still runs every day.

Start by listing running instances:

aws ec2 describe-instances \
  --filters Name=instance-state-name,Values=running \
  --query 'Reservations[].Instances[].{InstanceId:InstanceId,InstanceType:InstanceType,LaunchTime:LaunchTime,Tags:Tags}'

Then inspect CPU over the same review window the detector uses:

aws cloudwatch get-metric-statistics \
  --namespace AWS/EC2 \
  --metric-name CPUUtilization \
  --dimensions Name=InstanceId,Value=i-0123456789abcdef0 \
  --start-time 2026-03-09T00:00:00Z \
  --end-time 2026-03-23T00:00:00Z \
  --period 300 \
  --statistics Average

Look first for old instances with consistently tiny averages, no clear owner, and no obvious schedule requirement. Those are the best rightsizing or shutdown candidates.

Likely false positives and caveats

Low CPU is useful, but it is not the whole workload. Memory-bound databases, queue workers waiting for burst traffic, and maintenance hosts can all look quiet in CPU metrics while still serving a purpose.

That is why this detector is best used as a review queue. If you want infrastructure that is clearly stale rather than merely oversized, compare it with Idle Load Balancers and Idle RDS instances, which look for different kinds of weak activity.

How to fix underused EC2 instances

The safest cleanup path is:

1. Confirm workload ownership, maintenance windows, and rollback expectations.
2. Decide whether the instance should be stopped, scheduled, or resized.
3. Snapshot or verify attached data paths before major changes if the instance still carries unique state.

aws ec2 stop-instances --instance-ids i-0123456789abcdef0
aws ec2 modify-instance-attribute \
  --instance-id i-0123456789abcdef0 \
  --instance-type '{"Value":"t3.small"}'
aws ec2 start-instances --instance-ids i-0123456789abcdef0

How this differs from nearby AWS idle detectors

Idle NAT Gateways focuses on low-traffic network egress cost once that detector page is spec-backed and publish-ready. Idle Load Balancers looks for load balancers with no recent traffic or no targets. This detector is narrower: it only asks whether a running EC2 instance has stayed at very low CPU for long enough to justify a rightsizing review.

How Cloud Waste Hunter helps

Cloud Waste Hunter can turn low-CPU EC2 metrics into a practical compute review list with region, instance type, estimated savings, and enough context to decide whether to stop, schedule, or rightsize safely. For adjacent infrastructure cleanup, review the AWS Idle and Underused Resources guide.