Share this:
Your invoices match your forecast. That’s exactly why the leak is so hard to find.
Your infrastructure budget looks reasonable on paper. The invoices match the forecast, the renewals went through, leadership signed off. And somewhere between what you’re paying and what you’re actually getting, money is disappearing — most teams just never find out where.
The problem isn’t overspending in the obvious places. It’s the slow, quiet drain of costs that never make it onto a dashboard: systems that outlived their purpose, licenses that keep renewing, capacity provisioned for a peak that never came. These aren’t exotic edge cases. They’re structural patterns that show up in nearly every hybrid environment, and they compound — quietly, and faster than the budget line suggests.
The short answer
The ten costs below are not ten separate problems to chase down one at a time. They’re a single symptom wearing ten costumes: the inability to see your entire hybrid estate — cloud and on-premises — in one accurate picture. Hidden infrastructure costs are a visibility problem before they’re a spending problem. Which is why the move that actually compounds isn’t a cost-cutting initiative. It’s cost discovery before cost-cutting: a systematic effort to find what exists, what it costs, and whether it’s earning its place — before you decide what to kill. Until you have that baseline, every optimization is guesswork, and guesswork at the scale of a modern IT budget is expensive.
Why the leak keeps growing
Here’s the part that makes hidden costs worse than they look: the base they’re leaking from is rising fast. Gartner projects worldwide IT spending will exceed $6 trillion in 2026, growing nearly 10% year over year — and very little of that growth comes from organizations buying dramatically more technology. A fixed waste percentage against a growing base means the dollar value of the leak grows every year, even if you change nothing.
And the structural drag isn’t small. McKinsey’s 2026 analysis of enterprise technology spending found that the most disciplined operators — the ones it calls “deliberate modernizers” — keep the share of their budget tied up in run-based infrastructure costs at least 20% lower than everyone else, which frees that money for actual innovation. Everyone else is carrying the gap. The reason, per the same research, is that most organizations keep stacking new capabilities on top of legacy systems instead of retiring anything — so they pay twice, once to keep the old thing running and again to run the new thing on top of it, and the run cost only compounds. The takeaway McKinsey lands on is the one that matters here: in the AI era, it isn’t enough to spend more, you have to spend differently. On the cloud side, Flexera’s 2026 State of the Cloud Report — vendor research, so read it with that in mind — found that wasted cloud spend ticked back up to 29% after five straight years of decline, as AI workloads reintroduced cost complexity the discipline had only just gotten ahead of.
Rising base, persistent percentage, compounding leak. Here are the ten places it hides.
1. Zombie servers
These are servers that are technically running but serving no active workload. They draw power, cooling, and maintenance around the clock, and no one is using them — usually because no one is watching them. They tend to be the legacy of a migration that was 90% complete: the old environment never got fully decommissioned because the new one “wasn’t quite ready,” and then everyone moved on to the next fire.
Zombie servers are easy to miss precisely because they’re quiet. No tickets, no complaints, no alerts. Just a steady draw on the budget with nothing on the other side of the ledger.
2. Orphaned storage
Storage gets provisioned for a project, a workload, an application. The project ends, the application migrates, the workload retires — and the storage keeps running. In cloud environments these orphaned volumes accumulate fast, and they’re easy to overlook because they cause no visible problem. They just cost money, every billing cycle, indefinitely.
The same pattern lives on-prem in a different outfit: over-provisioned arrays, backup volumes from systems that no longer exist, snapshots that were “temporary” two years ago and have since taken up permanent residence.
3. Underutilized compute capacity
Hardware sized for peak demand spends the other 90% of its life running at a fraction of capacity. On-prem makes this especially common — a server gets bought for a worst-case scenario that rarely materializes, and the right-sizing conversation gets deprioritized because “we might need it someday.” Someday is doing a lot of work in that sentence.
The result is real cost attached to nothing: power, cooling, licensing, maintenance, and floor space, all supporting headroom that isn’t doing anything but waiting.
4. Forgotten software licenses
Licensing is where budget goes to hide. Enterprise agreements, seat-based SaaS subscriptions, per-server licenses — they share one trait, which is that they renew automatically and no one reviews them until something breaks. Employees leave and their seats keep renewing. Tools get replaced but the old contract runs out its term in the dark. Departments consolidate while their individual subscriptions don’t.
Across a large environment, the cumulative cost of software no one is actively using is rarely small and almost never tracked systematically. It’s the most clerical item on this list and one of the most expensive.
5. Unoptimized storage tiers
Not all data needs to live on high-performance storage. In a lot of environments it does anyway — hot storage becomes the default because tiering policies were never defined, or because moving data to a cheaper tier looked like low-priority effort at the time. The cost difference between tiers, especially in the cloud, is substantial. Parking infrequently accessed data on your fastest, most expensive infrastructure is one of the most reliable ways to overpay consistently and never notice.
6. Shadow IT and untracked cloud spend
The cloud’s greatest advantage — anyone with a credit card and an account can stand up infrastructure in minutes — is also one of its most expensive liabilities. Development teams, business units, and individual contributors provision resources outside formal procurement, and those resources frequently never get tagged, attributed, or terminated.
What you’re left with is spend that doesn’t appear in your cost-management tooling, isn’t tied to a business owner, and routinely outlives the project that created it. By the time finance asks the question, the answer is genuinely hard to reconstruct — not because anyone hid it, but because nothing was ever built to see it.
7. Network infrastructure nobody audits
Switches, routers, load balancers, firewalls — they get provisioned and forgotten. Unlike servers or cloud resources, network gear rarely triggers a cost conversation. It just runs. But the costs attached to it don’t sit still: maintenance contracts renew, support tiers get bumped “just in case,” and hardware sized for traffic that never showed up keeps drawing budget year after year.
Network infrastructure is also one of the least-audited categories in any environment. No ticket queue announces that something is over-provisioned. No alert fires when a support contract renews at a rate that stopped reflecting reality three years ago. The cost is real; it’s just invisible until someone goes looking.
8. Over-provisioned cloud resources
Right-sizing cloud workloads is harder than it sounds and far easier to defer than it should be. When a workload first ships, it gets provisioned conservatively — better to over-provision than to debug a performance issue at 2 a.m. The plan is always to right-size later, once things stabilize.
Later rarely arrives. The workload runs fine, nobody complains, and the oversized instance keeps billing at a rate higher than it needs to be. Multiply that across dozens or hundreds of workloads and the overage stops being a rounding error and starts being a line item.
9. End-of-life hardware still in production
Hardware past its supported lifecycle isn’t only a security risk — it’s a hidden cost. Vendors charge premium rates for extended support on EOL systems. Replacement parts get expensive and hard to source. Maintenance windows stretch. And the operational drag of keeping aging systems alive quietly absorbs engineering time that should be going toward higher-value work.
The business case for replacing EOL hardware is usually obvious on a slide. The organizational will to act on it is the harder thing to manufacture — especially when the system is stable and the risk hasn’t introduced itself yet.
10. Incomplete decommissioning
When a system is retired, decommissioning is rarely as clean as the runbook implies. The associated pieces — network configurations, monitoring agents, backup jobs, DNS entries, cloud security groups — tend to get left in place. Any one of them is trivial. Across a large estate and a few years of “we’ll clean that up later,” they add operational complexity, create security exposure, and carry real dollar cost.
The deeper issue is what incomplete decommissioning signals. If you don’t have a reliable process for retiring a resource completely, the odds are good your environment is holding more of these remnants than anyone can currently name.
The real problem isn’t any one of these
It’s that most organizations have no reliable way to see all ten at once.
Every item on this list is the same problem in different clothing: cost attached to something nobody can see clearly enough to question. That’s why chasing them individually doesn’t work — you fix the zombie servers, and the orphaned storage keeps billing; you reclaim the orphaned storage, and the forgotten licenses renew on schedule. The leak moves faster than a manual audit can follow.
This is the on-prem blind spot most cloud cost tools were never built to close. Flexera One, Datadog, CloudHealth, Apptio Cloudability — they’re capable platforms, and they stop at the cloud’s edge by design. That leaves the on-premises half of a hybrid estate — often the more expensive half, and the half where zombie servers, orphaned arrays, and EOL hardware quietly accumulate — running on quarterly spreadsheet estimates from a formula someone built three years ago. You can’t reconcile two halves of a budget when only one of them is instrumented.
Visual One Intelligence® closes that gap. It extends Hybrid FinOps™ discipline across the entire environment — cloud and on-prem in the same frame — and does the unglamorous work that makes a single view trustworthy: normalizing inconsistent tags into one business dimension, reconciling on-prem CapEx against cloud OpEx in a common language, and surfacing hybrid showback that business units actually believe because the numbers are measured rather than estimated. The result is that the question running underneath all ten of these costs — what exists, what does it cost, and is it earning its place? — finally has one accurate answer instead of ten partial ones scattered across tools that each see a fraction of the picture. That’s not a cost-cutting initiative. It’s the cost discovery that has to happen before any cost-cutting initiative can be more than a guess.
And the discovery tends to pay for itself faster than anyone budgets for. One airline had spent years trusting its capacity reports — the same reports every adjacent tool agreed with — before Visual One surfaced 730TB of hidden storage capacity and helped it avoid roughly $1.5 million in storage it was about to buy and didn’t need. The capacity wasn’t lost or broken. It was sitting in plain sight, invisible to every tool that only looked at part of the estate. That’s items #2, #3, and #5 on this list — orphaned storage, idle capacity, and mis-tiered data — found in one place, in one pass, because something was finally pointed at the whole environment at once. The other seven items don’t announce themselves either. They just need the same thing: a view wide enough to see them.
Questions IT leaders are asking about hidden infrastructure costs
What’s the difference between cost discovery and cost optimization?
Cost optimization is acting on what you can see — right-sizing an instance, retiring a license, consolidating a workload. Cost discovery is building an accurate, complete picture of what exists across your environment before you act, so optimization stops being guesswork. Discovery comes first because you can’t optimize what you can’t see, and most hidden costs are hidden precisely because no tool was ever pointed at them.
Why don’t existing cloud cost tools catch this waste?
Most FinOps and cloud cost tools were built for cloud-native environments and stop at the cloud’s edge. In a hybrid estate, that leaves on-premises infrastructure — servers, storage, network, and EOL hardware — outside the picture entirely. Hybrid FinOps closes that gap by normalizing cost and usage data across both cloud and on-prem into a single view, so the on-prem blind spot stops being where waste accumulates unseen.
How much of an IT budget typically goes to hidden or wasted costs?
The research points in one direction even when the exact figure varies. McKinsey’s 2026 analysis found that the most disciplined operators carry 20% less of their budget in run-based infrastructure costs than their peers — the rest of the field is spending that difference on complexity it can’t see. Flexera estimates 29% of cloud spend alone is wasted. The exact number varies by environment, but the pattern is durable enough that assuming yours is zero is the one safe way to be wrong.
You can’t cut what you can’t see
The IT leaders who make the best infrastructure decisions aren’t necessarily the ones with the biggest budgets. They’re the ones who know exactly what they’re working with — every server, every volume, every renewal, on both sides of the hybrid line. The leak isn’t a spending problem you can budget your way out of. It’s a visibility problem you have to solve first.
See your entire hybrid estate in one accurate view — and find the costs hiding in the half your current tools can’t reach.
