You're paying too much for AWS — here's how to fix it without a consultant

AWS is the world's most powerful cloud platform. It's also one of the easiest ways to burn money without realising it. Most teams that come to us for help aren't running bad infrastructure — they're just running infrastructure they forgot to turn off.

Why your AWS bill keeps climbing even when nothing changes

AWS charges for idle resources the same way it charges for active ones. An EC2 instance sitting at 2% CPU in a development environment costs real money every hour. So does the Elastic Load Balancer nobody decommissioned after last quarter's project. And the 500GB EBS volume still attached to a terminated instance.

The dirty secret of AWS cost management is that most waste doesn't come from bad decisions — it comes from forgotten ones. A team spins up a test environment, ships the feature, and moves on. Nobody deletes the environment. Three months later, you're paying $300/month for a ghost.

The good news: AWS gives you the tools to find this waste for free. You just have to know where to look and what to do with what you find.

The five-minute audit: where to start

Open AWS Cost Explorer and set the view to "Service" for the last 90 days. Look for anything that surprises you. EC2 is usually the top line — but look closely at EC2 "Other," which hides EBS, data transfer, and snapshots. These are where quiet waste accumulates.

Next, open AWS Trusted Advisor. Under "Cost Optimization," you'll see a prioritised list of idle resources, low-utilisation instances, and underused Reserved Instances. If you're on a Business or Enterprise support plan, you get the full list. If you're on Basic, you get enough to start.

The goal of this first pass is not to fix everything. It's to identify your top three waste sources. In most accounts, 80% of avoidable spend comes from three to five line items. Fix those first.

EC2: the biggest bill, the easiest wins

EC2 is almost always the largest AWS cost. The three fastest EC2 savings are: rightsizing, reserved instances, and stopping non-production environments overnight.

Rightsizing means switching from an instance type that's bigger than you need to one that fits your actual workload. If your m5.xlarge is running at 8% CPU, an m5.large will handle the same load at half the price. Use AWS Compute Optimizer — it analyses your CloudWatch metrics and recommends the right size.

Reserved Instances or Savings Plans are the single biggest lever for production workloads. A one-year Compute Savings Plan with no upfront payment typically saves 40% compared to on-demand pricing. If you have stable workloads — and most production systems do — this is free money.

Finally, schedule your dev and staging environments to shut down at night and on weekends. A $200/month development server that runs 16 hours a day instead of 24 costs $133. That's a 33% cut with zero performance impact.

S3, RDS, and the services people forget to audit

S3 costs creep up because teams set buckets to "Standard" storage class and never revisit them. Old log files, build artefacts, and backups from two years ago sit in Standard storage at $0.023/GB/month when they could sit in Glacier Instant Retrieval at $0.004/GB/month. Use S3 Lifecycle Policies to automate the transition.

RDS is expensive because databases are provisioned for peak load, not average load. If your RDS instance runs at 15% CPU most of the time with occasional spikes, consider Aurora Serverless v2 — it scales to zero during idle periods and only charges for capacity used.

Data transfer is the hidden cost nobody talks about. Moving data out of AWS to the internet costs money. Moving data between AWS regions costs money. Moving data between availability zones costs money. Audit your data transfer line in Cost Explorer and look for patterns that shouldn't be there.