The cloud bill almost always grows faster than expected, and it is rarely bad luck — it is a sign that no one is watching that spend with the same discipline applied to any other business cost. FinOps is not a one-off cut, it is a habit.
Where the waste usually is
- Underused resources — Instances and databases running 24/7 that are only needed during business hours, or test environments no one turned off.
- Rightsizing — Oversized machines "just in case", paying for capacity that is never actually used.
- Reserved vs. on-demand — Stable workloads paid at on-demand price when a 1- or 3-year reserved commitment would cut costs significantly.
- Cold storage — Historical data kept in the most expensive storage tier when it could move to infrequent-access tiers.
None of these measures require sacrificing performance — rightsizing often ends up improving performance by removing poorly-sized bottlenecks. What it takes is continuous visibility into spend, not an annual review.