The cloud promised paying for what you use. The reality three years in: you pay for whatever you once switched on and are now afraid to switch off. A third of a typical cloud bill is pure waste.
Where the money leaks
Dev and test environments running non-stop — used 40 of the week's 168 hours. Instances sized “to be safe” at ten times the real load. Database backups with infinite retention. Logs in premium storage that nobody has ever read. And orphaned resources from projects that ended two years ago.
Three steps in one afternoon
First: tag resources by project and owner — nothing works without that. Second: automatic shutdown of dev environments outside working hours, an instant 60–70 % saving on their cost. Third: a monthly per-project cost report to management; things someone looks at tend to slim down.
Reserved instances and spot: a discount for predictability
Paying full on-demand price for steady load you need all year anyway is like living in a hotel instead of renting. Reserved instances or savings plans cut that capacity by 30–60 %, and interruptible “spot” even more for suitable jobs. You just separate what must always run from what can wait — and buy accordingly.
FinOps is not a one-off
A one-time clean-up saves once; six months later the waste is back, because the team again “switched on and didn't switch off”. Only a rhythm works: a monthly per-project cost report, budget alerts and an owner who watches the numbers. The cloud is a rent that grows quietly — and it can be kept in check just as quietly.
Above €3,000 a month, professional optimisation virtually always pays off — we routinely find 25–40 % savings with no performance impact. That is the whole team's holiday, paid to Amazon for nothing.