Data centers & mission-critical

Why $/MW Replaced $/SF in Data Center Estimating

Two buildings with identical footprints can carry radically different price tags. Here is why capacity — not floor area — became the unit that mission-critical facilities are bought, benchmarked and built in.

By Sharib Maradukhel · MRICS · MCIOB · PQS · Director, Electrical Cost Management

For most of my career, the square foot was the honest shorthand of construction cost. You could look at a building's area, apply a benchmark from a comparable project, and land within a defensible range. On data centers, that shorthand quietly stopped working — and a great many budgets were embarrassed before the industry admitted it.

The reason is straightforward once you see it. A 100,000 square foot enterprise facility and a 100,000 square foot AI-optimized facility are not the same asset. One might run 5 to 10 kW per rack; the other, 40 to 130 kW. The building did not change. The electrical and cooling infrastructure did — dramatically. Benchmarking one against the other is not a small error with a percentage adjustment. It is a category error.

Electrical scope is the budget, not a line in it

On mission-critical projects, electrical systems routinely drive 40 to 50% of total construction cost — the largest single block in the plan. That single fact should reorganize how a cost team works. If your estimating process treats electrical as a percentage allowance derived from a building-wide rate, your contingency will not survive contact with the market.

It also means the electrical estimator is no longer a specialist consulted late. On these projects, the electrical single-line diagram is the cost plan. Everything else is quantities.

What capacity-based estimating actually requires

Moving to a cost-per-megawatt basis is not simply a change of denominator. It changes what you must know before you can price anything:

• Design density. The kW per rack assumption determines the electrical topology, the cooling strategy, and therefore the majority of the cost. It is the first question, not a detail.
• Redundancy tier. The step from concurrent maintainability to full fault tolerance duplicates entire electrical distribution paths — additional generation, more switchgear, more space, more commissioning. It is a structural cost decision, and it is frequently made in a sentence before anyone prices it.
• Cooling technology. Liquid cooling changes power distribution architecture, not just the mechanical scope. A benchmark drawn from an air-cooled hall will not survive a liquid-cooled design.

The gear is the schedule

The second discipline that mission-critical work forces on a cost team is procurement literacy. Transformers, medium-voltage switchgear and generators now set the critical path — not concrete, not steel, not weather. Lead times have stretched to the point where factory slots are secured before schematic design finishes.

The practical consequence for estimating is that a Class C estimate without a long-lead procurement plan is incomplete. Early equipment purchase changes the commercial structure of the project: owner-furnished equipment, letters of intent, and deposit-secured factory positions each shift risk, cash flow, and warranty exposure. Those consequences belong in the estimate, not in a surprise memo eighteen months later.

Labour became the second scarce commodity

Materials dominated the escalation conversation for years. On mission-critical projects, labour has quietly overtaken them. Electricians with data-center qualifications command significant premiums and are booked into builds months ahead of need. Commissioning specialists — the people who prove the facility actually works under failure conditions — are among the hardest roles in the market to secure.

This matters to the estimate in a specific way: wage growth in specialized electrical trades has decoupled from general construction labour indices. If your estimate escalates all labour at one blended rate, you are systematically underpricing the trades that matter most.

None of this makes area-based benchmarking useless. It makes it insufficient. Capacity, density and redundancy are the variables that move the number now, and a cost professional who cannot interrogate all three is pricing a building that no longer exists.