A meaningful share of cloud spend on a typical estate is recoverable without changing the architecture at all. Start there.
You cannot reduce what you cannot attribute. Before anything else, enforce a tagging policy so every resource carries an owner, an environment and a cost centre, and make untagged resources fail to create.
Then produce a breakdown by team and by service. The first run of this report is usually uncomfortable and always informative.
None of these require architectural change, and together they usually account for a substantial reduction.
Once the baseline is genuinely steady, committed-use discounts and reserved capacity offer meaningful savings for workloads that will certainly keep running. The mistake is committing before rightsizing, which locks in the waste.
Rightsize first, observe for a period, then commit to the reduced baseline.
One-off optimisation exercises decay. Costs creep back as new resources appear without tags, environments get cloned and forgotten, and retention settings default.
A short monthly review, attended by engineers rather than only finance, keeps the estate honest. Showing teams their own spend changes behaviour more reliably than any policy document.
Beyond cleanup, the durable savings come from architecture. Autoscaling that actually scales down, storage lifecycle rules that move data to cheaper tiers automatically, and serverless components for spiky workloads all reduce cost without ongoing effort.
Data transfer is the charge most often overlooked. Traffic between regions and out to the internet accumulates quietly, and architectures that move data unnecessarily can cost more in transfer than in compute.
Design decisions made with cost in view are far cheaper than optimisation applied afterwards, which is why engineers seeing spend data changes outcomes more than any policy.
It varies with how long the estate has been running unmanaged. Estates that have never had cost attribution usually find a significant proportion recoverable within a quarter through rightsizing and cleanup alone.
Native cost tooling is adequate for most organisations. Dedicated platforms add value at larger scale or across multiple providers, but they do not replace the tagging discipline that makes any of it useful.
Yes. Engineers make cost decisions constantly, and they make better ones when the consequences are visible. Hiding spend in a finance report guarantees it stays unmanaged.
Keep reading
The overspend is rarely caused by the cloud itself. It is usually set in motion by three decisions taken in the first few weeks, often by people who did not realise they were making them.
All four run the same workloads competently. The decision usually follows existing commitments, data residency and the skills already in your team.
Kubernetes is excellent at problems many organisations do not have. Here is a straight test for whether you are one of the ones that does.
