Where cloud waste actually hides

Every bill we review has the same shape. A small number of line items carry most of the waste, and none of them look alarming on their own.

Instances sized for a launch that already happened

Compute is usually the largest single block, and usually the most overprovisioned. Teams size for the worst hour of the worst week, then never revisit it. Utilisation sits in the teens for months while the bill stays flat.

Storage nobody has looked at

Logs and snapshots accumulate on the most expensive storage class by default. A lifecycle policy that moves anything older than thirty days to cold storage is often the cheapest change available, and the one nobody owns.

Traffic taking the long way round

Egress and gateway charges are quietly enormous when internal services talk to each other through the public path. Private endpoints usually pay for themselves in weeks.

Environments that never sleep

Development and staging run around the clock for a team that works office hours. Scheduled shutdowns recover most of that spend without touching production.

What to do first

Rank by realised savings, not by effort. The largest wins are almost always the least glamorous, and they are the ones a review should surface in days, not weeks.