Topic
Data transfer
Nobody budgets for data transfer, which is how it becomes the bill's most consistent surprise.
Data transfer pricing is the part of the AWS bill with the most moving parts: direction, source, destination, and path each change the rate. Traffic between availability zones costs money in both directions. A NAT gateway charges for every gigabyte it processes on top of its hourly rate. The same bytes leaving through CloudFront can cost less than leaving S3 directly.
Because transfer charges attach to other services' line items, they rarely get read as one number, and the architecture decisions that drive them are made by engineers who never see the bill. A chatty cross-AZ service mesh or a NAT gateway on a high-volume data path can outcost the compute behind it.
The pieces here map the transfer pricing matrix, the architectural patterns that generate surprise egress, and the fixes with real payback.
Related topics
Storage
S3, EBS, and archival storage. The classes, lifecycle rules, and hidden retrieval costs that decide the bill.
Compute
EC2, Lambda, and Fargate. Instance families, rightsizing, and the price of idle capacity.
Billing fundamentals
How the AWS bill is built. The CUR, rate types, cost allocation tags, and reading an invoice.