Budgeting rack power for colocation

Getting rack power right before you order colocation space avoids two expensive surprises: ordering more capacity than you need, or discovering mid-deployment that your equipment doesn't fit within what a rack, or a single power feed, can actually deliver. The methodology below is standard datacentre practice, not specific to any one provider, and it's worth working through before committing to a rack size or power allocation.

Step 1: sum the nameplate power of every device

Every piece of rack equipment, a server, a switch, a storage array, carries a nameplate or rated power figure, usually printed on the chassis or listed in its datasheet, expressed in watts or as a maximum amperage at a given voltage. Start by listing every device you plan to install and its rated draw, then sum them. This total is your starting point, not your final budget, it represents the theoretical ceiling, not what the equipment will actually draw under normal operation.

Step 2: apply headroom, don't plan to the theoretical maximum

Real-world power draw is almost always lower than the nameplate figure, nameplate ratings are typically set conservatively, well above what a device draws under normal, even heavy, load. But planning exactly to nameplate sum leaves no room for peak load, added equipment later, or measurement error, and running infrastructure consistently near its rated maximum leaves no margin for the unexpected. A common approach is to budget with meaningful headroom below the summed nameplate total, rather than treating that sum as the target to fill. Exactly how much headroom to build in depends on how variable your workload's power draw is and how much future growth you want to leave room for, so treat this as a planning principle to apply deliberately, not a fixed formula.

Step 3: understand what redundant feeds actually protect against

A rack fed by two independent power paths, commonly labelled A and B, is only meaningfully redundant if either feed alone can carry the full load on its own. If your actual draw is split evenly across both feeds but adds up to more than either feed can handle by itself, a failure on one feed doesn't gracefully halve your capacity, it overloads the surviving feed and risks taking the whole rack down at the worst possible moment, exactly when you needed the redundancy to work. The budgeting rule that follows from this: keep your total draw within what a single feed can carry alone, even though you're drawing from both. Dual power supplies on your own equipment, plugged one into each feed, are what let this actually work in practice, without dual supplies on the equipment itself, a feed failure means an outage regardless of how the rack is wired.

How this maps to what Worldstream publishes

Worldstream's colocation facilities run on N+1 redundancy with dual power feeds and support up to 10kW of power per rack as an available density option, details already covered in Power, cooling and rack density at Worldstream's data centres. The 10kW figure describes what a rack can be provisioned for, it isn't a target to fill, the same headroom-and-single-feed-capacity reasoning above applies to it directly. Facility-level efficiency, expressed as a PUE of around 1.15, is a useful conceptual reference too: it's a reminder that the power a facility draws to run your rack is somewhat higher than what your equipment itself consumes, cooling and distribution overhead account for the difference, though PUE describes the facility as a whole rather than any single rack.

A worked approach, in outline

1. List every device going in the rack with its rated power draw, and sum them.
2. Apply headroom below that sum to arrive at a realistic planning figure, rather than budgeting to the theoretical maximum.
3. Check that planning figure against what a single feed can carry alone, not the combined capacity of both feeds, since that's what stays available if one feed fails.
4. Confirm your equipment actually has dual power supplies wired one to each feed, otherwise the redundancy the facility provides doesn't reach your hardware.
5. Leave room in the plan for equipment you'll add later, rather than sizing exactly to what you're racking on day one.