Sizing rack power comes down to five steps: total the actual load rather than the nameplate ratings, convert that load into amps at your supply voltage, apply the continuous load rule that limits a circuit to 80 percent of its rating, divide across phases and across A and B feeds, then add headroom for growth. The step most often skipped is the 80 percent derate, which is why racks designed on paper to fill a 30A circuit trip in practice.
Step 1: Total the actual load, not the nameplate
Add up what the equipment in the rack will really draw, not what is printed on the back of it.
Nameplate ratings describe the maximum a power supply could theoretically pull, usually with every option fitted and every component at full tilt. Real draw is commonly a good deal lower. Designing to nameplate is the fastest way to strand capacity, because you will fill the rack on paper long before you fill it electrically.
Where to get better numbers: manufacturer power calculators for the specific configuration, measured draw from existing identical equipment, or metered data from a similar rack already in service. If none of those exist, use nameplate but record that you did, so the assumption can be revisited once real measurements are available.
A practical tip: measure over a realistic period rather than at a single moment. Idle draw and sustained-workload draw can differ substantially, and the number that matters is the sustained one.
Step 2: Convert load to amps
Circuits are rated in amps, and IT equipment is described in watts. You need to move between them, and the formula depends on whether the supply is single or three phase.
- For single phase: Amps = Watts / Volts
- For three phase: Amps = Watts / (Volts x 1.732)
The 1.732 is the square root of three, and forgetting it is a common error that overstates the current a three-phase circuit needs.
- Worked example, single phase. A rack drawing 5,000W at 208V single phase needs 5,000 / 208 = 24.0A.
- Worked example, three phase. The same 5,000W at 208V three phase needs 5,000 / (208 x 1.732) = 13.9A per phase.
That difference is the reason higher-density racks move to three-phase supply: the same power arrives over meaningfully less current, and therefore less copper.
Step 3: Apply the continuous load derate
This is the step that gets missed, and it is the reason racks that look fine on a spreadsheet trip breakers in service.
A load running at its maximum for three hours or more is generally treated as a continuous load, and electrical codes commonly require that a circuit carrying a continuous load be limited to 80 percent of its rating. Essentially all IT equipment qualifies as continuous.
In practice that means a 30A circuit gives you 24A of usable capacity. A 20A circuit gives you 16A.
Applying it to the example above: a rack needing 24.0A at 208V single phase does not fit on a 30A circuit, because that circuit’s usable capacity is 24A and there is no margin at all. It needs either a larger circuit or a three-phase supply.
Some rack PDUs enforce this internally by alarming or limiting at 80 percent of their rating. Do not rely on that; size for it deliberately.
Codes and their editions vary by jurisdiction. Treat the 80 percent figure as the working rule and confirm the requirement with the engineer of record against the code edition that applies to your site.
Step 4: Divide across phases and feeds
Two divisions happen at once, and they interact.
- Across phases. On a three-phase rack PDU, outlets are assigned to phase pairs. Loads should be spread roughly evenly, because a PDU fully loaded on one phase and lightly loaded on another will trip protection while apparently having capacity left. Most intelligent rack PDUs report per-phase current, which makes this straightforward to verify after installation and worth checking whenever equipment is added.
- Across A and B feeds. In a 2N arrangement each rack draws from two independent paths. The critical constraint: under normal operation each side should carry no more than half its usable capacity, so that either side alone can take the entire rack if the other fails.
That last point is the one that surprises people. A 2N rack fed by two 30A circuits does not have 48A of usable capacity. It has 24A, because each side must be able to carry everything alone. Sizing 2N as though the two sides add together removes the redundancy you paid for.
Step 5: Add headroom, then check the receptacle
Decide the growth allowance deliberately rather than by accident. Rack density has been rising steadily, and a rack sized exactly to today’s equipment will be full at the first refresh.
How much headroom is a judgement call tied to how disruptive adding capacity later would be. In a room where adding a circuit means a shutdown, headroom is cheap insurance. Where distribution can be extended live, less is needed.
Then check the physical connection, which is where designs most often come apart on installation day.
A quick worked example, start to finish
A rack with 20 servers, each measured at 400W sustained, in a 2N configuration at 208V three phase.
- Step 1. Total actual load: 20 x 400W = 8,000W.
- Step 2. Convert: 8,000 / (208 x 1.732) = 22.2A per phase.
- Step 3. Derate: at 80 percent, a 30A circuit provides 24A usable. 22.2A fits, with very little margin.
- Step 4. In 2N, each side must carry the full 22.2A alone, so both A and B feeds need that 30A three-phase circuit. They do not share the load in a way that reduces either side’s requirement.
- Step 5. With only 1.8A of margin, a single additional server would exceed the circuit. If any growth is expected, move to a larger circuit now rather than rebuilding the feed later.
This example is arithmetic, not a design. Any real installation should be reviewed by a qualified electrical engineer against the applicable code and the specific equipment involved.
Frequently asked questions
How do you convert watts to amps for a three-phase rack?
Divide the load in watts by the supply voltage multiplied by 1.732, which is the square root of three. For example, 5,000W at 208V three phase requires 5,000 / (208 x 1.732), or about 13.9A per phase. Omitting the 1.732 factor is a common error that overstates the required current. For single-phase supply, divide watts by volts with no additional factor.
What is the 80 percent rule for rack power?
Electrical codes generally require circuits carrying continuous loads, meaning loads running at maximum for three hours or more, to be limited to 80 percent of the circuit rating. Practically all IT equipment counts as continuous. So a 30A circuit provides 24A of usable capacity and a 20A circuit provides 16A. Code requirements and editions vary by jurisdiction, so confirm with the engineer of record for your site.
How much capacity does a 2N rack actually have?
The same as one of its two sides, not the sum of both. In a 2N arrangement each feed must be able to carry the entire rack alone so that the load survives losing either side. A rack fed by two 30A circuits therefore has 24A of usable capacity after the continuous load derate, not 48A. Sizing 2N as though the feeds add together removes the redundancy the design was intended to provide.
Should I size a rack to nameplate ratings or measured draw?
Measured or manufacturer-calculated draw wherever it is available. Nameplate ratings describe a theoretical maximum with every option fitted, and real sustained draw is commonly well below it. Sizing to nameplate strands capacity, because the rack fills on paper long before it fills electrically. Where no measured data exists, use nameplate but record the assumption so it can be revisited once the equipment is in service and metering is available.
Next step
If you want a rack layout or a power schedule reviewed before it goes to procurement, send it over. Trilogy designs and supplies rack-level power distribution and is happy to sanity-check the arithmetic.
