AWS Builder Center
How I Cut Our AWS Bill by 40%

How I Cut Our AWS Bill by 40%

NAT Gateway, idle RDS, and a Lambda hot loop were silently burning money. Three targeted fixes cut our AWS bill by 40% without any application code changes.

The Problem

Last quarter our AWS bill crept up 60% with no matching growth in traffic. Two hours in Cost Explorer identified three culprits: NAT Gateway data transfer at $800/month, oversized RDS instances at 8% CPU, and a Lambda calling S3 GetObject on every invocation instead of caching.

Fix 1: VPC Endpoints for S3 and DynamoDB

All outbound traffic was routing through NAT Gateway, including calls to S3 and DynamoDB. Adding VPC Endpoints for both services routes that traffic privately through the AWS backbone at zero data transfer cost.
Result: $800/month to $12/month.

Fix 2: Right-size RDS

We had db.r5.xlarge running at 8% average CPU. Peak usage never exceeded 35%. Moved to db.t3.large with GP3 storage.
Result: $680/month to $190/month.

Fix 3: Cache the S3 Config Object

A Lambda was fetching a config JSON from S3 on every cold and warm invocation. One module-level cache with a 5-minute TTL dropped request volume from 4 million to around 50 per day.
Result: negligible S3 cost vs $90/month.

The Pattern

Every cost audit I have run surfaces the same three areas: data transfer, idle compute, and chatty internal calls. Cost Explorer resource group breakdown plus a weekend with CloudWatch metrics is usually enough to find them all.
What is your biggest unexpected AWS cost reduction? Curious whether NAT Gateway charges are a universal surprise or just common in our setup.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article