🔪We cut AWS S3 costs by 40% overnight.
The storage wasn’t the problem.
A lot of companies assume their AWS bill is driven by infrastructure size.
In reality, it’s often driven by behavior.
In this case:
very small storage footprint
no heavy traffic
nothing “obviously expensive”
Yet the bill kept growing.
After digging in:
- ~85% of the S3 cost was data transfer
- caused by a pipeline re-downloading the same data every few seconds
Not a scaling issue.
Not a capacity issue.
Just inefficient usage.
We changed one parameter (polling frequency), and:
💥 costs dropped ~40% immediately
No migration.
No refactor.
No savings plans.