Cloud waste hit 29% of spend in 2026 — the review process that catches it before the invoice does

After five straight years of improvement, cloud waste went back up in 2026 — to roughly 29% of spend, from 27% the year before. On average, organisations are now losing about 32% of their cloud budget to idle resources, over-provisioning, and infrastructure nobody’s watching closely enough.

Some of that reversal has an obvious cause. 72% of organisations now use GenAI cloud services, but only 63% of FinOps teams are actually tracking that spend — which means roughly a third of the businesses paying for AI infrastructure have no systematic view of what it costs. Token-based billing and bursty inference workloads don’t behave like the steady-state compute most cost dashboards were built for.

Where the money actually goes

It’s rarely one dramatic mistake. It’s a staging environment that never got torn down, a database sized for a traffic spike that happened once, reserved capacity nobody re-evaluated after the last migration. Individually small, collectively enough that AI cost management is now the #1 priority for FinOps practitioners — cited by 98% of them in 2026, up from just 31% two years earlier.

The gap between organisations is stark. Businesses with a mature FinOps practice typically run at 10–15% waste; those without one see 35–40%. Same cloud provider, same services, very different bill — the difference is whether anyone’s actually reviewing the architecture on a schedule.

The review that keeps it under control

Structured cost optimisation programmes report an average 25–30% reduction in monthly cloud spend, and the mechanics behind that number are unglamorous: infrastructure as code so environments are reproducible and disposable, rightsizing on a recurring schedule rather than a one-off, and cost visibility that separates baseline infrastructure from bursty AI workloads.

None of it is exotic. It’s the same discipline that keeps deployments reliable — applied to the bill instead of the uptime.