Cloud cost teardown: the bill nobody was reading

A cloud bill that keeps growing while the traffic stays flat is almost never a usage problem. It is an unwatched line item problem.

The first pass on a subscription account usually finds the same four things. A development environment on production-sized instances, running around the clock because nobody built the scheduler. Block storage left attached to instances that were deleted months ago, still billed at the rate they were provisioned at. A gateway charged on traffic processed, fronting requests that a private endpoint would have carried for a fraction of it. And log retention set to never expire in every project, because the default was never changed.

None of those are dramatic read alone. Together they explain a monthly number that doubled without a single product decision behind it.

The fix is not a cleanup, it is a review the client can run themselves afterwards. Right-size from fourteen days of actual utilisation rather than the launch-day guess. Schedule the dev stack off outside working hours. Move shared services behind private endpoints. Put a retention policy on the logs and an alarm on month-to-date spend, so the next drift arrives as an alert in week one instead of a surprise at quarter end.

What a teardown should leave behind is a bill you can explain line by line, and an owner who can tell within a week when it starts to move.