How to Size Critical Power Infrastructure: A Modular UPS Story From the Emergency Desk
The call came in at 5:15 on a Friday
October 2024. I was wrapping up a week that had already included a 35kV transformer waiting on temperature-rise test results, a customs delay on a shipment bound for Houston, and a parameter sheet I still needed to sign off on. My phone buzzed.
An EPC contractor I'd worked with twice before. He was delivering a new electrical room for a regional data center in the Midwest. Their original modular UPS supplier had failed factory acceptance testing, and had been stalling on a retest for eleven days. The general contractor was already on-site. The raised floor was finished. They needed the UPS energized in three weeks.
Normal lead time: eight to ten weeks.
I'm a senior applications engineer at PowerSentinel. Over the past eight years I've handled more than 200 rush power infrastructure orders, including same-week turnarounds for data center and utility clients. I asked him to send the load schedule and put on another pot of coffee.
What the load schedule didn't say
If you've ever sized a UPS under deadline pressure, you know the load calculation is the first domino. Get it wrong, and everything downstream—switchgear, feeders, even the physical layout of the electrical room—is wrong too.
His spec called for a 600kVA modular system, six 100kVA modules. The original supplier had based it on 2022 server specifications with a 20% margin.
Then the GC said something that made me stop scrolling.
"We measured the actual load on-site. It's running about 340kW right now."
340kW. At 0.95 power factor, that's roughly 358kVA. Add standard headroom, and a 400kVA system would cover it.
I ran the numbers again. At N+1—which is the minimum for any data center claiming Tier III redundancy—a 500kVA modular system (five 100kVA modules, four running plus one spare) gives you 400kVA of usable load capacity. 358kVA fits comfortably inside that.
So why had the original supplier planned for 600kVA?
Because they used 0.9 power factor. 340kW divided by 0.9 is 378kVA. Add a 20% margin and you're at 450kVA. Then N+1 takes you to 600kVA—six modules, five running plus one spare.
The assumption that added a quarter million in cost
Here's the thing. 0.9 power factor was a reasonable assumption in 2015. Server power supplies back then were less efficient, and reactive power was a genuine concern. But a data center being commissioned in 2024? Modern IT loads run between 0.95 and 0.99 power factor. The GC's on-site measurements confirmed it.
That 20% gap between 0.9 and 0.95 isn't academic. It's the difference between five modules and six. At roughly $25,000 to $35,000 per 100kVA module—including the switchgear and busway connections—you're looking at a six-figure swing before you even factor in the extra floor space and installation labor.
The 1 a.m. decision
Honestly, I hesitated. If I recommended downrating the system and the actual load came in higher than projected, I'd own that failure. Nobody remembers the engineer who saved money on paper. They remember the guy whose UPS tripped on go-live day.
So I did what I always do on rush orders: I called the GC back and asked for the raw measurement data. Not the summary, not the screenshot—the actual logs. Then I walked through them with our applications engineer.
The 340kW reading was solid. It was taken on a Tuesday at 2 p.m.—peak load for a data center, with servers, cooling, and lighting all running. The measurement method checked out.
The only open question was growth. If their IT load doubled in three years, the 600kVA system would be justified. If it stayed flat, 600kVA was waste.
That's when the GC told me the client's IT expansion was phased. Phase 1—the current build—was live now. Phase 2 comes in 2026. Phase 3 in 2028. By 2028, total load is projected at 500kW.
At 0.95 power factor, 500kW is 526kVA. At that point, a 500kVA system (five modules, N+1) genuinely won't be enough. They'll need to add a sixth module, bringing it to 600kVA.
So the original 600kVA spec wasn't wrong. It was just four years early.
Our recommendation: start with 500kVA—five 100kVA modules in N+1 configuration. When Phase 2 comes online in 2026, add a module. No need to replace the frame, no need to rewire the room, no downtime. That's the whole point of modular architecture.
The GC ran the numbers and agreed.
Energized on day 14
We spent three days locking the technical spec, five days on production and FAT, two days on logistics, and the rest on-site commissioning. On day 14, the UPS was energized and load-tested successfully.
The GC later told me the savings roughly covered an upgrade to their monitoring system. But the point of this story isn't how fast we moved.
When modular UPS is the wrong choice
If you've read this far, you might be thinking modular UPS is always the smart play—flexible, scalable, future-proof. But here's the honest part: if your situation doesn't match, modular is the expensive choice.
My experience is based on about 140 critical power projects, mostly in the 200kVA to 2MVA range. If you're working below 100kVA, some of this won't apply to you.
Here's how to know if you're in the other 20%:
- Your load is stable and won't grow for at least five years. A small manufacturing facility, for example, where the equipment list is fixed and the load curve is flat. In that case, a monolithic UPS costs 15–25% less than a modular system of the same capacity, and maintenance is simpler. Don't pay for flexibility you'll never use.
- Your total capacity is below 100kVA. The economics of modular don't work at low power. The frame, busway, and control system overhead adds cost that you can't amortize across enough modules. A 100kVA monolithic UPS plus a battery cabinet will be cheaper, full stop.
- Your budget is hard-capped and your growth forecast is a guess. Modular's whole value proposition is paying for capacity only when you need it. If you're not sure you'll ever expand, that value is zero. A monolithic system at least lets you spend less today.
From the outside, modular UPS looks like it's always more expensive—extra modules, extra controls, extra enclosures. The reality is that if you have a genuine phased buildout, the lifecycle math flips in modular's favor by Phase 2. The problem is, most companies don't have a real phased plan. They have a hunch.
That's the distinction I keep coming back to.
What I'd tell the next client
This project made me rethink something: sizing isn't just about numbers. It's about timing.
The original 600kVA spec would have been correct in 2028. But installed in 2024, it meant paying for four years of capacity that did nothing but consume power, take up floor space, and require maintenance.
Modular UPS lets you buy later. But that only works if your expansion plan is real. If you're just "thinking you might grow," modular isn't your friend—it's an excuse to overspend.
And one last thing: get the power factor assumption right before you do anything else. The gap between 0.9 and 0.95 isn't a rounding error. On a 500kVA system, it's the difference between compliant and non-compliant. On a 600kVA system, it's a module you didn't need to buy.
Fix the input assumptions first. Everything else follows.