How to Reduce Your AWS Bill: 9 Practical Steps That Actually Work

Most AWS bills carry 20–30% of waste that nobody chose on purpose — it accumulates from forgotten resources, default settings, and on-demand pricing for steady workloads. The good news: a handful of changes recover the bulk of it. Here are nine steps, ordered roughly by effort versus payoff.

1. Delete unattached EBS volumes

When you terminate an instance, its data volumes don't always go with it. Orphaned volumes in the available state bill 24/7 for storage you're not using. Find them, confirm nothing needs the data (snapshot first if unsure), and delete. At roughly $0.08/GB-month for gp3, a few stray 100GB volumes is real money.

2. Release unassociated Elastic IPs

AWS now charges for public IPv4 addresses. An Elastic IP that isn't attached to a running resource costs about $3.60/month each — trivial individually, but they pile up across accounts and regions. Release the ones you don't need.

3. Clean up old snapshots and AMIs

Snapshots are incremental, but years of daily backups with no lifecycle policy add up. Set a retention policy (for example, keep 30 days), and delete snapshots and deregistered AMIs that no longer map to anything you run.

4. Migrate gp2 volumes to gp3

gp3 is about 20% cheaper than gp2 per GB and lets you provision IOPS and throughput independently. For most volumes it's a cheaper, faster, drop-in replacement you can change live. See our dedicated guide on gp2 vs gp3.

5. Right-size over-provisioned instances

Check CloudWatch CPU and memory utilization over a few weeks. Instances that sit under ~10% CPU are candidates to drop a size (halving the size roughly halves the cost). Move to current generations too — newer instance families usually give more performance per dollar.

6. Stop or schedule non-production environments

Dev, staging, and QA rarely need to run nights and weekends. Scheduling them off for ~12 hours a day and weekends can cut their compute cost by 60–70%. Note: stopped instances still bill for their attached EBS storage, so terminate the ones you truly don't need.

7. Cover steady usage with Savings Plans

Any compute that runs around the clock on on-demand pricing is overpaying. A 1-year, no-upfront Compute Savings Plan typically cuts that baseline by ~30% with full flexibility across instance families and regions. We compare the options in Savings Plans vs Reserved Instances.

8. Watch for cost anomalies

A sudden jump — a runaway job, a misconfigured auto-scaling group, a chatty NAT gateway — can double a bill before anyone notices. Set up anomaly alerts so spikes surface in hours, not at month-end.

9. Allocate costs back to teams

You can't cut what you can't see. Tag resources by team, service, and environment, then show each team their spend. Accountability does more for long-term cost than any single cleanup.

Where to start

Steps 1–4 are low-risk and often recover the most in the first week. CloudWireHub runs all of these checks automatically against a read-only connection and shows you the dollar impact of each — start free and see your own numbers in minutes.