How to Cut Your AWS Bill by 40% Without Touching Performance

#The Average Scale-Up Wastes 35–50% of Their Cloud Spend

We've audited dozens of AWS environments for Series A and B companies as part of our cloud architecture consulting. The average waste we find is 38%. Not because engineers are careless — because cloud infrastructure grows organically and nobody has time to clean it up. This is one of the first things a fractional CTO addresses: getting visibility into where money is going and stopping the bleed.

Here's the systematic approach we use.

#Step 1: Get Visibility First

You can't optimize what you can't see. Start with:

• AWS Cost Explorer — enable it if you haven't, tag everything by service/team/environment
• Compute Optimizer — AWS's own tool that flags over-provisioned EC2, RDS, and Lambda
• Trusted Advisor — free tier gives you basic rightsizing recommendations

Spend one week just tagging and categorizing before touching anything.

#Step 2: Kill the Obvious Waste

In every audit we do, these are the first things we find:

Idle resources: EC2 instances running at <5% CPU for weeks. RDS instances nobody queries. Elastic IPs not attached to anything (you pay for those).

Over-provisioned databases: A db.r5.2xlarge running a dev environment. A Multi-AZ RDS setup for a staging database. These are common and expensive.

Forgotten snapshots and AMIs: Old EBS snapshots accumulating for years. Each one costs money. Run a cleanup script monthly.

Data transfer costs: Moving data between availability zones, regions, or out to the internet adds up fast. Audit your NAT Gateway costs — they're often the biggest surprise.

#Step 3: Rightsize Compute

Once you have visibility, rightsize in this order:

1. EC2 → Graviton: Moving from x86 to ARM-based Graviton instances gives you 20–40% better price/performance with zero code changes for most workloads.

1. Reserved Instances / Savings Plans: If you have predictable baseline load, commit to 1-year Savings Plans. Typical savings: 30–40% vs on-demand.

1. Spot for non-critical workloads: Batch jobs, CI/CD runners, dev environments — all good candidates for Spot. 70–90% cheaper than on-demand.

#Step 4: Optimize Storage

S3 is cheap but not free at scale:

• Move infrequently accessed data to S3 Infrequent Access or Glacier
• Enable S3 Intelligent-Tiering for data with unpredictable access patterns
• Review and tighten S3 lifecycle policies

For EBS: audit volume types. gp2 volumes should be migrated to gp3 — same performance, 20% cheaper, and you can tune IOPS independently.

#Step 5: Cloud Architecture Changes for Long-Term Savings

The biggest savings come from cloud architecture changes — this is where engineering leadership pays for itself:

• Serverless for spiky workloads: Lambda + API Gateway is dramatically cheaper than always-on EC2 for workloads with variable traffic
• CloudFront for everything public: Reduces origin load and data transfer costs
• RDS Proxy: Reduces database connection overhead, lets you use smaller instances
• AI-powered monitoring: Use agentic AI tools to continuously scan for cost anomalies and auto-recommend optimizations

#What to Expect

A typical 3-month engagement on a $50k/month AWS bill:

• Month 1: Tagging, visibility, kill obvious waste → 15–20% reduction
• Month 2: Rightsizing, Reserved Instances → additional 15–20%
• Month 3: Architecture optimizations → additional 5–10%

Total: 35–50% reduction, sustained.

This kind of process transformation — from reactive cloud spending to strategic cloud architecture management — is exactly what a fractional CTO engagement delivers. The savings typically pay for the engagement itself within the first month.