AWS Cost Optimisation: 7 Patterns That Cut Our Clients' Bills by 40%

Why AWS Bills Spiral Out of Control

AWS billing is notoriously opaque. You provision resources in a sprint, ship the feature, and move on. A year later the architecture looks nothing like the original plan and nobody has reviewed whether the infrastructure still fits the actual workload.

After auditing AWS environments ranging from $8K to $240K monthly, the over-spend patterns are remarkably consistent. Here are the 7 that drive the largest reductions.

Pattern 1: Right-Size EC2 Instances

The most common waste we find: EC2 instances running at 5–15% average CPU utilisation because they were sized for a peak workload that never materialised. AWS Compute Optimiser analyses your actual utilisation and recommends right-sized instances. The average recommendation saves 20–30% on EC2 costs alone.

Pattern 2: Reserved Instances and Savings Plans

On-Demand pricing has a massive premium over Reserved Instances. For any workload running more than 8 hours per day, 1-year Reserved Instances or Compute Savings Plans deliver 40–60% savings. Most clients avoid commitment because they're "not sure about future requirements" — but even a 70% commitment covers baseline load and dramatically reduces the bill.

Pattern 3: S3 Intelligent-Tiering

S3 Standard pricing for data that is accessed less than once per month is pure waste. Enable S3 Intelligent-Tiering on buckets containing data older than 30 days — it automatically moves infrequently accessed data to cheaper tiers with no performance impact for data accessed less than once per month. Average saving: 40% on S3 storage costs.

Pattern 4: NAT Gateway Elimination

NAT Gateway charges data processing fees of $0.045 per GB. For applications transferring large volumes of data through NAT Gateway (e.g., Lambda functions downloading S3 objects), this becomes a major cost centre. The fix: use VPC endpoints for S3, DynamoDB, and other AWS services — eliminating NAT Gateway charges for those data paths entirely.