Chapter 5 — drawn from Optimizing Corporate Efficiencies by Arthur J. Riel

Most large organizations are quietly bleeding millions of dollars a year on infrastructure decisions nobody has questioned in years. Not because anyone made a bad call — but because nobody made a new one. Contracts roll over. Servers stay racked. Mainframes keep humming. And the budget keeps absorbing the cost of decisions that were right five or more years ago and wrong today.

I've seen this pattern play out across infrastructure, data centers, and legacy systems throughout my career, and the fix is almost always the same: stop assuming the status quo is still the best option and go check.

The $12M Lesson in Never Skipping the RFP

While managing the infrastructure and middleware department at a major global bank, I inherited a four-year, $20M storage contract that was up for renewal. The team's plan was simple: renew with the existing vendor, just like the prior three cycles. That's what had always been done.

I insisted on a competitive RFP instead, expecting maybe a 5–10% price reduction from the threat of competition. What came back instead was a relatively new entrant in the storage market with a fundamentally different approach — and a bid at 40% of the incumbent's price. In a commodity category like storage, that kind of gap is almost unheard of.

After due diligence, we made the switch and saved over $12M across the four-year contract. The lesson is not "this vendor was bad." It's that infrastructure categories — storage, servers, networking, UPS systems — see real innovation cycle after cycle, and the winner from one generation does not automatically stay the winner in the next. If you're not forcing a competitive bid on every renewal, you're leaving money on the table by default, not by decision.

The Servers Nobody Remembered They Had

Data centers present a different kind of inertia problem: physical infrastructure with long leases and expensive, long-life assets (power, HVAC, redundant networking) that make consolidation feel risky and disruptive. I've led two major data center restructurings, and both surfaced the same uncomfortable discovery.

The first project consolidated five data centers into two, outsourcing them to a data center vendor and cutting costs by 45%. But the more interesting finding came at the start of the project, when my team needed to inventory the server stock and the workloads on each. They shockingly reported that it looked like 3–5% of servers (150 to 250 servers) were sitting idle — racked, powered, and cyclically replaced, but doing nothing. By the end of the program the actual number was 22%, or 1,100 servers.

This happens because there's rarely a clean process for decommissioning servers. A team requests a server, puts a system into production, eventually retires it — and nobody pulls the plug, because the infrastructure team is afraid of disrupting something they cannot easily verify is dead. Multiply that across a few thousand servers and you get a meaningful chunk of your budget paying for nothing.

The second restructuring — moving a single on-campus data center to a vendor site ahead of a looming $20M power and HVAC upgrade — found 8% zombie capacity and dropped total cost of ownership by 30%, not even counting the avoided capital expense. Even if you have no interest in restructuring your data center footprint, a server usage audit alone is worth the exercise.

The Mainframe Everyone Avoids

Mainframes are the most stubborn form of infrastructure inertia, mostly because most data center vendors will not take them, and full decommissioning is expensive and slow. In one case, the best proposal I received to eliminate a mainframe and rebuild its nightly batch processing on Linux came in at $18M with a three-year timeline. That price tag kills most decommissioning projects before they start.

Instead, we outsourced the mainframe to a vendor that specializes in mainframe consolidation. The transition took a few months, cost 35% less than running it in-house, and improved performance. The batch window, which had been running tight against its 8-hour limit, dropped to under two hours. We could even reduce the maximum MIP count, something that is not really possible to monetize when you own the mainframe yourself, opening the runtime window further and saving close to $1M in additional costs.

It's worth noting that AI-driven code translation is starting to make the more ambitious option — full mainframe decommissioning — economically realistic for organizations that have been sitting on these projects for years because the ROI never quite worked. That math is shifting.

The Pattern Underneath All Three

None of these savings required new technology or a creative breakthrough. They required someone to stop and ask, "is this still the best option?" instead of letting the contract renew, the server stay racked, or the mainframe stay untouched by default.

Infrastructure inertia isn't a technology problem — it's a governance problem. Build in the habit of forcing competitive bids, auditing actual usage, and revisiting "too hard to touch" assets on a regular cycle, and you will find millions in savings that have nothing to do with cutting corners and everything to do with simply checking your assumptions.

Next in this series: Cloud migration is not free — the decisions nobody tells you about.

Continue reading: You Are Wasting 10 to 45 Percent of Your IT and BPO Budget · Infrastructure & Cloud Transformation Services