Wiser, Not Thinner: Why Across-the-Board IT Cuts Cost More Than They Save

Share this:

Posted in:

Business Transformation, FinOps & Cost Management

The 10% memo isn’t strategy. It’s symmetry. And there’s a math that beats it.

By Noah Duke, with insights from Tom Mack, Chief Technologist, Visual One Intelligence

Across-the-board IT budget cuts are the most expensive way an IT organization can choose to save money, and the executives who use them know it. McKinsey has been documenting this since 2009, and the survey data is consistent across fifteen years of cycles: the executives who run blanket cuts admit, in their own words, that they doubt the savings will hold.¹ They keep doing it anyway, because the alternative is harder. The alternative, when you have the right view of your infrastructure, is also the way to find more savings without losing anything that’s working — and that’s what this post is about.

The trap of fairness

The reason organizations default to across-the-board cuts isn’t that nobody read the McKinsey memo. It’s that across-the-board cuts solve a political problem more than a financial one. When the CFO walks in asking for 10%, the path of least organizational resistance is to make every department bleed equally. The alternative path requires walking into a room and explaining that the database team is going to take a bigger hit than the network team, with workload-level evidence to back the call. That conversation is uncomfortable on the way in, harder on the way out, and impossible without data most organizations don’t actually have — particularly on the on-prem side of the environment, where most FinOps tools simply don’t look.

So they default to fairness. Fairness is what you do when you don’t have evidence.

The math that beats the memo

There’s a better way to find that 10%, and it doesn’t involve cutting 10% from anything.

“The normal thinking is ‘I need to save 10% this quarter.’ Everything becomes an accumulated incremental change, and nobody’s looking for the moonshot. With visibility, what they’re doing is — I need to save 5% over here, 3% over there, 7% over there.”
— Tom Mack, Chief Technologist, Visual One Intelligence

The math compounds. Five percent off compute, three percent off storage, seven percent off legacy systems, two percent off duplicate tooling, and that’s seventeen percent in aggregate, none of it coming out of capability that was actually working. The blanket 10%, by contrast, takes its full bite out of load-bearing capacity right alongside the waste, which is a different way of saying that the blanket 10% saves nothing the executive intended to save and quietly damages the things the executive intended to keep.

The FinOps Foundation’s 2026 State of FinOps report frames the same pattern from the practitioner side: mature teams report diminishing returns from traditional optimization, having already captured the “big rocks” of waste and now facing “a high volume of smaller opportunities that require more effort to capture.”² The discipline of the next decade is the discipline of compounding small wins, and it’s available only to organizations that can see at the right level of granularity to find them.

The numbers in the field are typically more dramatic than this framing suggests. Most enterprise virtualization footprints we’ve reviewed are running at 22 to 27% utilization when, with the right operational discipline, they could safely run at 50 to 70%. That’s not a 10% optimization opportunity. It’s a doubling of effective capacity sitting unused inside infrastructure the business has already paid for, and nobody is calling it that because nobody can see it at the level of granularity required to call it anything.

Why the math isn’t available to most teams

The reason most IT organizations can’t make this kind of targeted call has very little to do with skill, and almost everything to do with structure. What makes hybrid cost discipline hard is organizational, not technological. The VMware administrator can see the VMware footprint at the workload level, the storage administrator can see the SAN at the workload level, and the network team can see the network at its own level of granularity, but no single role, including the manager whose job is to actually make the budget call, has a normalized view of how those layers add up to a workload, an application, or a business outcome. The view exists across seven different consoles, and reconciling them is a job nobody is paid to do.

The competing incentives inside the silos make the problem worse. Managers want high utilization because high utilization is efficient use of capital; system administrators want low utilization because low utilization is what keeps the on-call phone quiet at 2am. The result is a quiet equilibrium where nobody has the whole picture, every conversation happens at the wrong level of abstraction, and the budget gets cut symmetrically by default because that’s the only call anyone has the data to defend.

The fix isn’t a framework. It’s the kind of normalized view Visual One Intelligence® was built to provide — a split-screen across VMware, Nutanix, Hyper-V, on-prem storage, and cloud, with automated tag normalization translating “Env:Prod” and “System:Production” into a single business dimension so a manager can ask data-driven questions of the specialists who own each layer without first having to log into seven consoles. The data is normalized, the workloads are attributable, and the hybrid showback is finally defensible. That visibility is what unlocks the 5%-here, 3%-there compounding math. Without it, the only defensible move is the blanket cut, and the blanket cut is what got us here.

A seven-figure illustration

A recent example: a customer early in a Visual One proof of concept told us they couldn’t fund the platform because they’d just committed to a seven-figure storage expansion. The data, after running for a single day, told a different story. The environment had a substantial amount of orphaned capacity, including storage that had been allocated but never reclaimed and orphaned LUNs scattered across the SAN. The customer didn’t believe the numbers, went back to verify, and confirmed it: the orphaned capacity was real, and the storage expansion wasn’t necessary.

The pattern matters more than the dollar figure. The organization had already paid for the capacity. They simply couldn’t see it. Without seeing it, the only available move was to buy more, which is exactly the kind of decision that feels conservative on the procurement side and turns out to be the most expensive call in the entire stack.

The 2026 stakes

This argument is getting more urgent, not less, and contrary to most of the headlines, the urgency isn’t coming from where you’d expect.

“I thought the AI cost conversation would be the center of this. It’s not. The cost conversation in 2026 is centered around memory and flash prices. And it’s weird because there’s no extra value in that. There’s a ton of extra value in AI.”
— Tom Mack

Memory and flash prices have gone through the ceiling, with no relief expected until new fabs come online in 2028. Gartner’s April 2026 forecast pegs data center systems spending growth at 55.8% year-over-year, attributing the surge to AI infrastructure demand alongside record price increases in high-bandwidth memory.³ The CFO mandate to find 10% will become more aggressive over the next eighteen months, not because IT teams are wasteful, but because hardware unit economics are working against every infrastructure leader at once.

The defense isn’t a smaller budget; it’s sweating the resources the business has already paid for. In Visual One customer environments, the typical result is running roughly 60% more virtual machines on existing hardware, deferring server purchases, and stretching refresh cycles from four years to five or six. That’s not a 10% conversation. It’s a “you don’t have to buy the next round of servers at all” conversation, and the only way to have it is with the kind of hybrid visibility that surfaces capacity at the workload level — precisely what Visual One was built to deliver across the hybrid footprint cloud-only FinOps tools don’t see.

Questions IT leaders are asking

Why do across-the-board IT budget cuts often fail?

They treat every dollar of spend as equally optional, which means they cut load-bearing capacity at the same rate they cut waste. The savings are often eaten by outages, emergency over-purchases six months later, or missed migration windows, a pattern McKinsey has documented for over a decade. The executives who use blanket cuts admit, on the same surveys, that they doubt the savings will hold.

What’s the difference between cutting IT costs and optimizing IT spend?

Cutting reduces what the business pays. Optimizing reduces what the business pays for things that aren’t producing value, while protecting what is. The first is a number on a memo. The second requires data on what every workload actually costs at the infrastructure level, which most organizations have on the cloud side and almost never have for on-prem — the gap Visual One Intelligence is built to close with normalized tagging and hybrid showback across the full footprint.

How quickly can a hybrid IT cost optimization effort show measurable savings?

First measurable savings typically show up within a quarter. The compounding effects, including retiring legacy workloads, rebalancing capacity, and deferring hardware purchases, usually accumulate over months six through twelve. In Visual One deployments, the conversation shifts within the first review: organizations stop arguing about percentages and start arguing about specific systems.

The unsexy version is the version that works

The plain version of this argument is the one that lands hardest:

“We report on things that aren’t sexy. But if you do the things that aren’t sexy well, that makes the business.”
— Tom Mack

Wiser, Not Thinner is unsexy by design. Not a moonshot, not a transformation initiative — just the discipline of finding 5% in five places, defending the spend that’s actually working, and refusing to fund what nobody can defend with data. That’s how you find capacity sitting in a storage array nobody had checked, how you avoid buying servers you don’t need, and how you beat the 10% memo with math.