The three patterns behind cloud cost inflation
We have reviewed cloud estates for dozens of enterprises across banking, manufacturing, and services. The cost-inflation patterns are remarkably consistent across sectors:
Orphaned resources. VMs, databases, storage volumes, and load balancers that were created for specific projects and never decommissioned when the project ended. Typically 8-18% of a mature estate runs on resources that no current system actively depends on.
Right-sizing gaps. Workloads provisioned at peak-load capacity running at 20-30% average utilisation because nobody has the data or the mandate to rightsize them. The original provisioning was typically done by the team that built the workload, under deadline pressure, and never revisited.
Commitment mismanagement. Most hyperscale vendors offer 20-50% discounts on committed capacity (reserved instances, savings plans) — but only for capacity actually consumed. Enterprises without a FinOps discipline either under-commit (leaving discount money on the table) or over-commit (paying for capacity they don't use).
Why the IT team cannot solve this alone
The underlying problem is a principal-agent mismatch. The team provisioning cloud resources is not the team that sees the bill. The team that sees the bill does not have the operational context to challenge specific provisioning decisions. The result is that cost visibility and cost accountability sit in different organisations, and the gap between them is where inflation lives.
FinOps, done properly, closes that gap. It creates a shared operating model where finance sees consumption in business terms (dollars per business unit, per product line, per customer), engineering sees cost implications of their provisioning decisions at the point of decision, and leadership sees a structured view of the tradeoff between cloud investment and business outcomes.
Without this operating model, every cost-reduction exercise becomes a fire drill that produces short-term savings and then slowly reverts. With it, cost optimisation becomes continuous and compound.
What a FinOps programme delivers in year one
Our typical FinOps engagement follows a four-phase structure — assess (4 weeks), design the operating model (4 weeks), implement tooling and governance (8 weeks), optimise continuously (ongoing).
Year-one outcomes typically include a 25-35% reduction in run-rate, establishment of cost showback or chargeback to business units, a commitment strategy saving an additional 15-20% on committed capacity, and — most importantly — a sustainable operating model where engineering decisions carry cost context and finance sees cloud spend in terms it can report to the board.
The cost savings are usually the easy part. The harder part is the cultural change required to make engineers care about cloud cost as a design consideration. That part takes longer, but it is the difference between one-time savings and compound value.
Where to start
For enterprises that have not yet formalised FinOps, we recommend starting with a 4-week structured assessment. The deliverables: a current-state cost map by business unit and service, a prioritised list of cost-reduction opportunities (typically 15-25 concrete actions), a commitment-optimisation recommendation, and a proposed operating model for ongoing FinOps discipline.
The assessment typically self-funds — the opportunities identified usually exceed the cost of the assessment by 10-20x on year-one impact. It is one of the rare infrastructure investments where the CFO is the likely sponsor, not the CIO.