The phrase that does the most damage to the average office IT installation is six words long, sits on an architect's drawing somewhere near the lobby, and reads: "Small server room, 6 by 8 feet."
I have walked into rooms exactly like this in cities across India. The wall-mounted electrical panel is neatly labelled. The split-AC unit is humming. There's a fluorescent tube on the ceiling. Two racks sit against one wall, packed with switches, the firewall, the camera NVR, the access control server, the wireless controller, and a UPS that someone bought because the brief said "with battery backup." Everything is plugged in. Everything appears to work. And every summer, almost without fail, something in the room overheats and brings down a system that the business depends on.
The problem isn't the equipment. The problem is that the room was designed by people who calculate cooling loads for kitchens and meeting rooms — not for the small, sealed, electrically dense space that an IT room actually is. A server room is not an electrical room. Treating it as one is one of the most common, most expensive, and most preventable mistakes in commercial IT.
Why this keeps happening
The chain of decisions usually goes like this. The architect lays out the floor plan and notes "small server room" as a back-of-house space, often near the electrical riser, sized by aesthetic instinct rather than calculation. The MEP team picks up the brief and provides what their training tells them to provide for a small back-of-house room: lighting, a power circuit, a smoke detector, an air conditioner sized to the room's volume. The IT contractor arrives at handover, fills the room with equipment, and does their best to make it work.
Nobody in this chain is wrong, exactly. The architect drew a room. The MEP team specified it correctly for what they were told it was. The IT contractor installed the gear they were paid to install. But nobody in the chain asked the question that matters: how much heat will this room generate, and how much power will the equipment in it draw, when it's running at capacity on a Tuesday afternoon in May? If you don't ask that question early, you cannot answer it later cheaply.
The five things a server room actually needs
1. Cooling sized for the equipment, not the room
A standard split-AC unit in a small office room is sized by floor area: so many square feet, so many tons. That math is wrong for a server room because the heat load doesn't come from the sun, the windows, or the people. It comes from the equipment. A modest two-rack setup — a core switch, a couple of distribution switches, a small NVR, a UPS, and the assorted appliances that an SME accumulates — generates somewhere between 5,000 and 12,000 BTUs of continuous heat. That's roughly the cooling load of a small living room, packed into a sealed box of fifty square feet.
The fix is not "bigger AC." The fix is the right kind of AC: a precision cooling unit, or at minimum a wall-mounted split sized for the calculated thermal load, ideally with redundancy. The room should be sealed, the airflow should be intentional, and the cooling should run continuously — not switched off at 8 p.m. when the office closes. A server room that overheats overnight will be a server room with shortened equipment life and unexplained Monday-morning failures.
2. Electrical capacity calculated, not assumed
The single most common UPS sizing error I see is "five minutes of backup." Five minutes is enough time to make tea. It is not enough time for the storage array to flush its write cache, the database to commit its open transactions, the application servers to gracefully shut down, and the network equipment to power off cleanly. The right number for a small SME setup is closer to fifteen to thirty minutes — not because you'll run on battery for that long, but because you need a window long enough to either ride out a brief outage or shut down gracefully if it persists.
Sizing the UPS correctly means knowing what's plugged into it, what each device draws under load, and what the diversity factor looks like. None of this is exotic. All of it requires that someone be in the room with both an electrician's understanding of power and an IT engineer's understanding of equipment behaviour. That person rarely exists on a typical office fitout brief.
3. Hot and cold aisles, even in small rooms
Hot/cold aisle layout is usually associated with large data centres, but the principle scales down. The basic idea is that all equipment in a rack pulls cool air in from the front and exhausts hot air out the back. If your two racks are arranged so that one rack's exhaust feeds directly into the other rack's intake, you've built an oven inside your server room. The fix is trivial — orient the racks so they face the same way, or place them back-to-back with the cooling supply directed at the front of both — but it has to be planned at room layout, not retrofitted later.
4. Monitoring you can actually see
Every server room should have, at minimum, four things being monitored: temperature (at least one sensor, ideally two — at the top of the warmest rack and at the AC return), humidity, UPS status, and water leak detection on the floor. All four can be logged to a small network appliance for under fifty thousand rupees, and the alerts can be routed to whoever's responsible for the IT estate. It is astonishing how many server rooms have none of this. People discover the air conditioner has been failing for three days only when they walk past and notice the door is warm.
5. Fire suppression that doesn't make the problem worse
The ceiling-mounted water sprinkler that comes as standard with most commercial fitouts is exactly the wrong fire suppression for a server room. It will, in the event of a fire, ruin every piece of equipment in the room — including the equipment that wasn't on fire. The correct answer for a small IT space is a clean-agent system: FM-200, Novec, or an inert gas, depending on budget and code requirements. These systems suppress fire by displacing oxygen or absorbing heat without leaving residue or damaging equipment. They cost more than a sprinkler. They cost vastly less than a total room replacement and a week of downtime.
What "right" looks like, briefly
A properly designed two-rack room for a typical mid-sized SME isn't expensive, and it isn't complicated. The room is sized to allow at least one metre of clearance in front of and behind the racks for service access. The cooling is calculated to the actual equipment load, with a redundant unit if uptime matters. The power is dedicated, separately fused from the main building load, and runs through a UPS sized for orderly shutdown. The racks face the same direction, with cool air entering at the front and hot air exhausted at the back into the AC return. A sensor cluster reports temperature, humidity, power, and water status to whoever needs to know. The fire suppression is a small clean-agent system, properly serviced. There's a single-line diagram on the wall showing what's plugged into what, and the cabling is labelled.
That description is not luxury. It's the minimum viable design for a room that businesses actually rely on. It costs more than the version that gets installed by default. It costs much less than the version that gets installed by default, multiplied by the number of times it fails over its lifetime.
What architects and project managers can do
Three things will save more pain than almost anything else you do on an office fitout:
Ask for the thermal load before sizing the room. The IT contractor or the ELV consultant can give you a rough number based on the equipment list. That number changes everything: room size, cooling spec, electrical capacity, cable tray sizing. Ask before the room is drawn, not after.
Specify the cooling, UPS, and monitoring as part of the brief — not as "by IT contractor." Leaving these line items off the construction package is what produces the "small electrical room with two racks" problem. Put them in the BoQ. Specify the cooling capacity in BTUs or kilowatts, not in tonnage. Specify the UPS in runtime minutes, not in kVA alone. Specify the monitoring as a deliverable.
Bring in someone who has actually sized one before. The skill of designing a small server room properly is not difficult, but it is specialised. It draws on cooling load calculation, power distribution, network architecture, and equipment behaviour at the same time. Most MEP teams won't do it; most IT contractors will defer to the electrical consultant. The right person on the brief is someone who has done it before and knows what failure looks like.
The cost of getting it wrong
The room continues to work for the first year. Maybe two. Then a switch fails, and the post-mortem reveals it ran fifteen degrees above its rated operating temperature for the previous eighteen months. Or a backup runs at 2 a.m. while the AC is off, the room hits 45°C, and the storage array's RAID rebuild fails halfway through. Or the UPS battery, which has been on a slow voltage decline that nobody monitored, finally dies in the middle of a presentation to the board.
Every one of these scenarios produces a remediation cost that exceeds, by a comfortable margin, what it would have cost to design the room correctly in the first place. The math on prevention vs. remediation is almost always one-sided. The reason it doesn't get done is that the prevention spend happens at design stage, and the remediation cost happens later, at someone else's spending head. By the time you're paying for the new switch, the architect has been paid, the MEP team is on the next project, and the IT contractor is being asked to "make it work."
If you're scoping an office fitout right now, or planning one in the next twelve months, the room labelled "small server room" on your floor plan is one of the cheapest things you can over-engineer and one of the most expensive things you can under-engineer. Half an hour at the design stage with someone who knows what to ask is almost always the highest-leverage conversation in the project.