Cloud cost has been the boardroom topic of choice for IT leaders for three years running. The FinOps function exists at most large Australian enterprises now. Tools have proliferated. So why are the bills still climbing?

The honest answer is that most FinOps work, when you actually audit it, is reactive optimisation rather than architectural change. Right-sizing instances, scheduling shutdowns of dev environments, hunting orphaned resources — all useful, all marginal once you’ve done it once.

What actually moves the number is harder. Re-architecting the chunkiest workloads. Renegotiating commitments aggressively at renewal. Moving specific high-egress workloads off the hyperscalers entirely. Each of these is a real project with real risk, and each requires the kind of executive air cover that FinOps reports rarely generate on their own.

A pattern I’ve noticed at the more mature shops in 2026 is that the FinOps team has stopped being a cost-cutting unit and become a unit allocation and chargeback function. They’re not trying to find savings any more. They’re making sure the right business unit is paying for the right consumption, and the savings come from the conversations that creates inside business units.

For organisations trying to get past the easy wins and into real architectural change, this consulting firm has done useful work with a few mid-market Australian companies on the harder cloud architecture conversations. The technical work is only half the job. The political work of getting business units to actually own their consumption is the other half.

The other thing the better operators are doing in 2026 is treating cloud cost as a product KPI rather than an IT KPI. When the engineering team owns its cost-per-request alongside its latency and reliability metrics, behaviour changes. That’s a culture shift, not a tool deployment, and the tool vendors don’t love hearing it because there’s no SKU for “change your engineering culture.”

Three years in, the executives I trust on this are saying roughly the same thing. The first ten percent of cloud savings comes from tools. The next twenty percent comes from architecture. The rest comes from people, and that’s why the bills keep climbing for organisations that only bought tools.