The State of FinOps in the Hybrid Enterprise: Why "Cloud Cost Management" Is the Wrong Frame
The term "cloud cost management" implies that the problem being solved is a cloud problem. The tools in the category were built that way — ingesting cloud billing APIs, analyzing cloud resource utilization, optimizing cloud commitments. For the first decade of FinOps as a discipline, that framing was close enough to true.
It isn't anymore.
The infrastructure that modern enterprises actually run spans cloud providers, on-premises datacenters, colocation facilities, containerized workloads, AI inference services, and a growing inventory of software contracts and vendor commitments. The cost management problem isn't a cloud problem. It's an infrastructure problem — and "cloud cost management" has become a category label that systematically excludes most of what it needs to cover.
The organizations that are ahead of this shift have stopped asking "how do we manage our cloud costs?" and started asking "how do we manage all of our infrastructure costs?" The difference isn't semantic. It changes the tools they use, the conversations they have, and the decisions they're equipped to make.
How We Got Here
FinOps as a formal discipline emerged from a specific problem: the public cloud's variable billing model created costs that were unpredictable, opaque, and growing. Infrastructure teams that had spent decades managing fixed capital expenditures suddenly faced metered consumption that could multiply in days without anyone noticing.
The FinOps Foundation, formed in 2020, codified the practices that had emerged organically among cloud-heavy organizations — a framework built around the cloud billing model, the cloud resource taxonomy, and the cloud optimization levers that existed at the time.
That framework was the right response to the problem as it existed then. The problem has since changed shape significantly.
Flexera's 2026 State of the Cloud Report found that 73% of organizations operate hybrid cloud estates — meaning nearly three-quarters of the organizations FinOps teams serve are running material workloads outside the public cloud, on infrastructure that most FinOps tooling treats as out-of-scope or handles as a secondary feature at best.
Meanwhile, AI spend has become a distinct and rapidly growing infrastructure cost category that doesn't fit neatly into the cloud billing model. Kubernetes has abstracted infrastructure in ways that break the assumptions underlying cloud-native cost attribution. Software contracts and vendor commitments represent billions in organizational spend that most FinOps practices don't touch at all.
The scope has expanded. The frame hasn't.
The Four Gaps in "Cloud Cost Management"
Gap 1: The datacenter. For most enterprises, on-premises infrastructure isn't a legacy remnant on its way to the cloud. It's a deliberate, ongoing operational choice for workloads where the economics, compliance requirements, or performance characteristics of public cloud don't make sense. Approximately one-fifth of enterprise workloads were repatriated from public cloud to on-premises environments in 2025 alone — a signal that the datacenter is getting more relevant, not less.
Yet the datacenter has no billing API. There's no Cost Explorer equivalent for your physical servers. Most FinOps platforms built for the cloud have no meaningful answer to the question "what does it cost to run this workload in our colo?" That gap leaves a significant fraction of enterprise infrastructure unmanaged by the tools and practices that are supposed to manage it.
Gap 2: AI infrastructure. AI spend isn't just growing — it's a new kind of cost that existing frameworks weren't designed for. Token-based pricing, multi-provider experimentation, inference spend that accumulates per API call rather than per instance-hour — these are billing models that don't map to the cloud cost patterns that FinOps tooling was built around. Enterprise GenAI spending grew from $11.5 billion in 2024 to $37 billion in 2025 — a 3x increase in a single year — and most FinOps practices have not kept pace with the new spend categories this growth has created.
Gap 3: Kubernetes. Container orchestration abstracts infrastructure in ways that break the traditional cost attribution model. With 96% of organizations now using Kubernetes, this isn't a niche concern. The cost of running containerized workloads is real and significant. The ability to attribute that cost to the teams and applications that generate it requires tooling and practices that sit above the cloud billing layer — which is where most "cloud cost management" platforms stop.
Gap 4: Software contracts. FinOps conversations are almost exclusively about consumption-based infrastructure spend. But enterprise IT budgets also include substantial software licensing costs, vendor agreements, SaaS subscriptions, and enterprise contracts with renewal obligations. These costs are managed (or mismanaged) separately — in procurement systems, in spreadsheets, in calendar reminders about renewal dates — without the same visibility and governance infrastructure applied to cloud spend. The total cost of operating a business's technology stack isn't visible in any single place.
What the Right Frame Looks Like
The organizations managing infrastructure costs effectively in 2026 have moved to a broader frame: infrastructure cost management, not cloud cost management.
Infrastructure cost management covers every dollar spent to run technology workloads — cloud, on-premises, containers, AI services, and the software and contracts that support them. It asks not "how do we reduce our AWS bill?" but "how do we get the most value from every dollar we spend on infrastructure, wherever that infrastructure runs?"
This reframing has practical implications for how FinOps teams operate.
It makes hybrid decisions possible. When you can see cloud costs and datacenter costs in the same view, with consistent attribution, you can make rational decisions about where to run workloads. Repatriation analysis, cloud migration cost modeling, buy-vs-build-vs-lease decisions — all of these require a unified view that cloud-only tooling can't provide.
It brings AI spend under governance. Treating AI infrastructure as a first-class cost category — with the same attribution standards, anomaly detection, and optimization practices applied to cloud compute — closes one of the fastest-growing visibility gaps in enterprise IT.
It creates accountability across the full stack. When every infrastructure cost category is attributed, tracked, and reported consistently, cost accountability conversations don't stop at the edge of what the cloud billing API can see. Engineering teams, finance teams, and business unit leaders have a complete picture — not a cloud picture with everything else in a separate conversation.
It positions FinOps as a strategic function. A FinOps practice limited to cloud billing optimization is a cost center with a narrow mandate. A practice that covers total infrastructure cost management has a mandate as broad as the organization's technology spend — and the ability to influence every significant infrastructure investment decision.
The FinOps Foundation's State of FinOps research shows the discipline maturing rapidly, with organizations increasingly moving past crawl-stage cloud cost visibility toward more sophisticated optimization and governance. The next maturity step, for most organizations, isn't doing cloud cost management better. It's expanding the scope of what "cost management" means.
The cloud was the beginning of the story. It was never supposed to be the whole story.
Reduce was built for infrastructure cost management — cloud, datacenter, Kubernetes, AI, and software contracts — in a single unified view. Because the cost of running technology doesn't stop at the cloud provider's billing boundary.