Six reasons your cloud bill keeps growing (and what to do about each)

Nobody makes one expensive mistake

When we are asked to look at a cloud bill that has grown uncomfortably, we almost never find a single bad decision. We find twenty reasonable decisions, each defensible at the time, none of which anyone has revisited since. Cloud spending grows quietly because the platform makes it trivially easy to add capacity and mildly inconvenient to remove it.

Here are the six causes we encounter most often, roughly in order of how much money they typically account for.

1. Resources sized for a worst case that never arrives

Virtual machines and databases are commonly provisioned against a peak estimate made before anyone had real usage data, and then left there. It is not unusual to find production servers running consistently below twenty per cent utilisation. Right-sizing against actual observed usage, with headroom, is usually the single largest saving available and carries very little risk when done carefully.

2. Environments nobody turned off

Development, test, training and proof-of-concept environments are created for a purpose and rarely deleted once that purpose passes. Worse, they typically run twenty-four hours a day for work that happens during office hours. Scheduling non-production environments to shut down outside working hours removes roughly two thirds of their cost immediately, and deleting the genuinely abandoned ones removes all of it.

3. Storage that only ever grows

Storage is cheap per gigabyte, which is precisely why it accumulates unchecked. Old snapshots, orphaned disks from deleted machines, backup copies retained far beyond any policy requirement, and log data kept indefinitely all add up. Applying lifecycle rules, moving older data to cheaper tiers and deleting what policy says should be deleted, is usually straightforward and often substantial.

4. Paying on-demand rates for predictable workloads

If a workload runs continuously and will continue to, on-demand pricing is the most expensive way to buy it. Reserved instances, savings plans and committed-use discounts trade flexibility for a significant discount. The mistake is committing to the wrong shape, which is why this exercise should follow right-sizing rather than precede it.

5. Licences that do not match reality

This is particularly common with Microsoft 365. Licences remain assigned to people who left months ago. Everyone is on the same tier regardless of whether they need it. Add-ons are purchased separately that are already included in a bundle the organisation holds. A licence review is a short piece of work with a reliably good return.

6. No ownership of the bill

The underlying cause of the other five. When the cloud invoice is treated as a fixed overhead paid by finance rather than a variable cost owned by someone technical, nobody is looking at it. Assign an owner, tag resources so costs can be attributed to a team or a project, set budget alerts, and review monthly. This costs nothing and prevents the problem from recurring.

A sensible sequence

Work through it in this order: get visibility and tagging in place; remove what is genuinely unused; right-size what remains; apply lifecycle policies to storage; review licences; and only then commit to reserved capacity. Committing first locks in whatever waste already exists.

A cloud cost review is typically a contained piece of work, a week or two, and in most cases the savings identified exceed the cost of the review by a comfortable margin. If your bill has been going in one direction for a while, it is worth having someone look at it.