You can order four racks, have them delivered, and get them cabled before lunch. Whether they stay online depends on a breaker you probably haven't looked at in a year.

Colocation capacity problems rarely announce themselves. They show up at 2 a.m. as a tripped branch circuit, a UPS running warmer than it should, or a redundancy setup that quietly stopped being redundant months ago. The audit below is the work that prevents that. Do it before you place the order, not after the racks are bolted down.

Start With the Paper Trail, Then Verify It on the Floor

Pull the one-line diagram, the rack elevation, and the circuit schedule for your cage. Find the committed power figure in your colocation contract. That number, not the physical capacity of the building, is often your real ceiling.

Then walk the floor and check the labels. A panel schedule from two years ago may not match a breaker that was renamed during a maintenance window, and rack labels migrate every time a cage gets rearranged. Ten minutes with a flashlight saves a deployment.

Three questions to settle early:

• Is each circuit dedicated to your cabinets, or shared inside your footprint?
• Which circuits are actually in use, and which are labeled as spares?
• Where does each PDU feed originate — the same panel, or different ones?

If you can't answer those, you don't have a baseline. You have a rumor.

Measure Actual Load, Not Nameplate Load

Nameplate ratings describe worst-case draw on a good day. Real consumption depends on CPU utilization, fan curves, drive activity, and ambient temperature. The gap runs in both directions: older gear often pulls less than its label, while equipment added without a change ticket pulls more than the spreadsheet says.

Start where instrumentation already exists. Metered rack PDUs report per-outlet and per-branch numbers, and most expose them over SNMP. Where metering doesn't exist, a qualified technician can clamp the branch circuits at the panel.

Log for at least a week, ideally across a full month. You want peaks, not averages:

• Nightly backup and replication windows
• Patching and vulnerability scanning cycles
• VM migrations during maintenance
• Monthly jobs — reporting, image builds, database maintenance

Record kW, kVA, and per-phase amps separately. Three-phase circuits are notorious for imbalance, and the heaviest phase trips first.

Don't forget the load nobody budgeted for

Idle servers still draw meaningful power, and so does everything around them: KVMs, monitoring appliances, an old switch left patched "just in case," inline test gear, a laptop cart plugged into a cabinet PDU. All of it sits on your circuits and none of it appears in most capacity plans.

Read the UPS the Way the UPS Reads It

The UPS display gives you two numbers that matter: output power in kW and load as a percentage. On many units that percentage is based on kVA, not kW. A UPS with a 0.9 output power factor can reach its kW limit while the display still reads comfortably under 100%. Watch only the percentage, and you can be closer to the edge than you think.

Then check the architecture:

• Is the UPS a single block or a modular frame with several power modules?
• Is it configured N+1, N+2, or 2N?
• What is the static bypass rated for, and will the load survive a transfer?
• After an outage, how much extra load does battery recharge add?

That last point catches people. The UPS carries the load through an outage, then spends hours recharging while IT load is back at full. If you were at 85% before the event, you won't be after it.

Redundancy Math Sets the Real Ceiling

Capacity isn't the number on the door of the UPS room. It's what remains after the worst single failure.

Say a modular UPS holds four 250 kW modules in an N+1 arrangement: 750 kW of usable capacity with all modules healthy, plus one module in reserve. If your load sits at 700 kW, losing a module leaves 750 kW against 700 kW of demand. Technically online, practically fragile — no allowance for recharge, inrush, or growth.

For a 2N design the arithmetic is harsher. Each side must carry the entire load alone, because that is exactly what happens during maintenance. Measure both feeds. Dual-corded racks pulling 70% from the A side and 30% from B are not balanced, and that imbalance wastes capacity on both sides. A sensible working target is to keep post-failure load below roughly 70–80% of remaining capacity, so recharge and normal swings still fit. Whatever number you choose, write it down.

Check Breakers and Branch Circuits Before the UPS

The UPS can have room and a breaker can still trip. Branch circuits are the real choke point.

In North America, continuous loads are generally limited to 80% of the breaker rating. Rough working figures — confirm them with a licensed electrician for your specific installation:

• 20A, 208V single-phase: about 3.3 kW of continuous load
• 30A, 208V single-phase: about 5 kW
• 30A, 208V three-phase: roughly 8.6 kW

Adding racks usually means adding circuits, and the panel schedule may be the binding constraint rather than the UPS. Confirm with your provider that spare breaker positions and poles actually exist. Also check whether any existing circuit is shared: a rack that adds 1.5 kW to a circuit already near its derated limit will cause problems long before the UPS notices anything.

Document the Headroom You Actually Have

Build the table before you deploy. At minimum, capture:

• Rack or cabinet ID and location
• Circuit ID, source panel, breaker size, and phase
• Measured peak amps per phase, with the date measured
• Derated capacity and current headroom in amps and kW
• A-side or B-side feed, and whether the rack is dual-corded
• Which UPS, PDU, or busway feeds the circuit
• Notes on anything shared, undocumented, or uncertain

Then total it. You want two numbers: how many kilowatts you can add while keeping full redundancy, and how many before you cross your working limit. The second number answers the question that actually matters — four racks or two?

Before You Sign the Work Order

Get confirmation in writing. Ask your provider to state the available headroom on your committed circuits and the UPS configuration serving them. A verbal reassurance from a technician during a walkthrough is not a capacity plan.

Photo: Brett Sayles / Pexels