Cloud waste is systematic, predictable and preventable. What actually causes it, what it costs the business, and the six-step discipline that brings it back under control.
Cloud adoption has delivered on much of its promise: agility, innovation, elastic capacity on demand. It has also produced a quiet crisis. Behind the dashboards, most organisations are haemorrhaging money without noticing.
Let us look at the what, the why and the how.
29% of cloud spend is wasted, and in 2026 that figure rose for the first time in five years.
Nearly a third. Not of the IT budget — of the cloud bill specifically. And the number went up in a year when almost every organisation already had cost optimisation as a stated priority. That combination is the story: the waste is not caused by inattention. It is caused by the fact that nobody can see it clearly enough to act.
The classic pattern is an instance sized for a peak that never arrives, then left running around the clock because nobody owns the decision to stop it. Alongside it sits a long tail: volumes detached months ago and never deleted, snapshots kept because nobody is confident they are safe to remove, snapshot chains pinned in place by an obsolete image nobody uses, load balancers with no targets, buckets nothing reads.
None of it is dramatic individually. Collectively it is the bulk of the waste, and it accumulates silently because a resource that does nothing also raises no alerts.
Cloud billing is genuinely hard to read. Instance families, regions, commitment terms, storage tiers, egress and inter-region transfer all price differently, and the invoice arrives after the money is spent. Teams provision against what they understand, which is rarely the cheapest correct option.
Organisations frequently “move to the cloud” by moving virtual servers onto cloud infrastructure. It is the fastest route, and it is the most expensive one, because it carries every inefficiency of the old environment across the boundary and then adds a margin.
Cloud platforms are built to provide services. Want a web server? Spin one up. Need storage? Spin some up. Those services are usually far cheaper than running a virtual machine to achieve the same result.
I have had many IT leaders tell me it would cost far more to host their systems in the cloud. When I assess those systems, the cost is substantially lower once the workload moves to services rather than to virtual machines.
It is more work up front. Source-controlled infrastructure as code is your friend here — it makes the rearchitecting repeatable, and it keeps you out of vendor lock-in.
Resources provisioned outside any governance process go unmanaged and unbilled to anyone in particular. Tags are missing or wrong, so cost cannot be attributed. Ownership is unclear, so nobody switches anything off. Forgotten infrastructure keeps charging indefinitely.
Reduced return, slower innovation. Money spent on idle capacity is money not spent on building anything. When the cloud bill grows faster than the workload, the case for the next investment gets harder to make.
Finance and operations in firefighting mode. Cost overruns pull IT and finance into monthly reconciliation arguments instead of planning. The teams best placed to fix the structural problem are the ones with the least time to look at it.
Escalating risk. Sprawl and misconfiguration are a security problem as much as a financial one. Unowned resources are unpatched resources, and they expand the attack surface without appearing on anyone’s inventory. Untracked cost also undermines transparency — which auditors and regulators do notice.
| Symptom | What is actually happening | What to do about it |
|---|---|---|
| Idle and oversized instances | Sized for a peak, never revisited; running 24/7 with no owner | Continuous rightsizing, scheduled start and stop, delete orphaned volumes and snapshots |
| An unreadable bill | Cost cannot be attributed because tags are missing or inconsistent | Tag compliance rules, cost pools, showback and chargeback |
| Overprovisioning | No guardrail between a request and a running resource | Provisioning constraints, resource quotas, time-to-live enforcement |
| Budget overruns | Discovered at invoice time, weeks after the spend | Forecast-aware budgets, anomaly detection on daily spend |
| Manual cost control | Someone remembers to check, until they are on leave | Automate the checks and the shutdowns |
Inventory and categorise all cloud spending. Use real-time discovery to catalogue every service and the cost attached to it. Tag by project, team or business unit — you cannot allocate what you cannot identify.
Adopt continuous rightsizing. Set policies to resize oversized instances, stop idle resources on a schedule, and reclaim unused storage and unused reserved capacity. Automation is the point here: manual reviews neither scale nor persist.
Implement showback or chargeback. Translate consumption into a number each business unit recognises as theirs. Transparency is what produces accountability; nothing else reliably does.
Enforce budget guardrails. Define alerts when spend crosses a baseline, and use automation to stop provisioning at a threshold rather than reporting the breach afterwards.
Commit where usage is predictable. Reserved instances and savings plans reduce unit cost meaningfully for steady-state workloads. The prerequisite is knowing which of your workloads actually are steady-state, which is a coverage-analysis problem before it is a purchasing one.
Build a FinOps culture. Shared ownership across engineering, IT and finance. Cost metrics in team KPIs and sprint rituals, not only in the monthly finance pack.
Each of those steps maps to something specific in the product rather than to a general capability.
Discovery and allocation. Cost Explorer across owner, pool, service, region, account and day, with resource inventory across twelve first-class resource types and twelve cloud and data platforms. Cost pools typed as budget, business unit, team, project, CI/CD or asset; assignment rules with nine condition types that keep ownership current automatically; and virtual tagging computed independently of cloud-native tags, so allocation does not depend on every engineer having tagged correctly.
Twenty-six named optimisation checks. Not “potential savings identified”. Abandoned instances, images, load balancers, S3 buckets and Kinesis streams. Obsolete images, IPs, snapshots and snapshot chains. Instance rightsizing and underutilised RDS detection. Instances stopped but not deallocated. Volumes long unattached. Instance generation upgrades and cross-region migration opportunities. Reserved Instance and Savings Plan purchase opportunities. Short-living instances flagged as spot and preemptible candidates. Kubernetes rightsizing and an object-storage duplicate finder. Every check carries its own thresholds, pool exclusions and account skips, so it fits your environment instead of nagging about it.
Automated shutdown. VM power schedules start and stop instances without anyone remembering to.
Governance that fires before the invoice does. Expense anomaly detection against a rolling daily average, expiring and recurring budget policies, resource count anomaly detection, resource quotas, tag compliance covering required tags, prohibited tags and correlation rules, resource time-to-live with automatic lifecycle enforcement, and total and daily expense limits per resource or pool. Constraint violations are kept as history you can hand to audit.
Showback and chargeback. Pool-based, with forecast-aware overspend states — so a pool that is on track to breach is flagged before it does.
An open cost model. A dedicated collector ingests the FinOps Open Cost and Usage Specification (FOCUS), and cost data can be queried directly against it. Your cost model stays portable and auditable rather than locked inside a vendor schema. Raw billing exports to external BI, and reporting runs on a schedule.
Security signals in the same place as cost. Inactive IAM users, unused console access and open security groups surface alongside the spend, because sprawl is one problem with two invoices.
Across cloud environments under management, Certero sees an average cloud cost saving of 38%.
The conclusion is unambiguous: cloud overspend is systematic, predictable and preventable. Too many organisations treat it as a cost of doing business rather than a fixable problem.
The method is proven:
Apply that discipline and cloud stops being a runaway cost centre and starts behaving like a managed asset.
So: is your cloud spend uncontrolled, or under continuous discipline? If you are ready to remove the blind spots, book a demo and see the twenty-six checks run in a fully populated environment.
Other posts covering the same ground.
CerteroX was announced in San Diego as one platform bringing cloud cost management and SaaS optimisation together. Here is what was said at the show, and what the platform actually ships today.
FinOps was built for the cloud and stops at the data centre door. ITAM knows the assets and licences underneath. Neither is complete alone — and hybrid cost control needs both working from the same data.
Audit frequency went up, not down, as software moved to subscription. Meanwhile a fixed share of cloud spend is wasted every month. Both are real, and only one of them is a certainty.
Everything argued above is checkable. Name the publisher, the billing account or the platform you would argue with, and the session is built around it — the reasoning attached, not a summary slide.
No gated download at the end of it.