We Cut Our AWS Bill by 60% in 90 Days — Here's Exactly What We Did

When our CTO told me we were spending $47,000 a month on AWS and wanted that number cut significantly, I assumed we’d need months of painful architecture work. We didn’t. Ninety days later we were at $18,800 — a 60% reduction — and we hadn’t touched a single customer-facing feature.

Step 1: The Audit (Week 1-2)

Before changing anything, understand what you’re paying for and why. AWS Cost Explorer with resource tagging is the starting point. We found three categories of waste immediately: idle resources nobody knew were still running, dramatically over-provisioned compute, and data transfer costs nobody had ever looked at.

The idle resources alone — EC2 instances, RDS instances, load balancers — accounted for 22% of our bill. We killed them all in a week.

Step 2: Rightsizing (Week 3-6)

Most AWS accounts have compute running at 20-30% average utilization. Ours was at 18%. We used AWS Compute Optimizer recommendations, validated them against actual performance data, and downsized systematically. This cut another 25% of the remaining bill.

Step 3: Reserved Instances and Savings Plans (Week 7-8)

Once we knew our actual compute footprint, we committed to reserved capacity for the baseline. This alone cuts costs 30-40% vs on-demand for steady-state workloads. We’d been putting this off for two years because it felt like a big decision. It took three hours.