Buyer's guide

What a data center PDU is, and how to specify one

12 August 2026 9 minute read

"PDU" covers four distinct pieces of equipment — a floor-standing PDU, a remote power panel, a rack PDU and a busway tap-off — each sitting at a different point between switchgear or UPS output and the IT load. The type, monitoring grade, phase configuration and breaker coordination all have to be specified separately. A comparable quotation depends on stating all four.

Which tier of PDU you are actually choosing between

"PDU" is an umbrella term for four physically distinct products, positioned at different points along the complete distribution chain from utility entry to rack.

A floor-standing (room) PDU is a free-standing cabinet with its own step-down transformer, rated from around 300 kVA upward, sitting between switchgear or UPS output and the room. Some floor PDUs integrate a static transfer switch, letting the unit accept two diverse upstream sources rather than one.

A remote power panel (RPP) sits downstream of a floor PDU or a stand-alone transformer, with no transformer of its own. A rack PDU, also called a cabinet distribution unit (CDU), is the last piece of equipment before the IT gear's own power cords, typically mounted at the rear of the rack. A busway tap-off draws from bus bars run overhead or under a raised floor.

The tier choice is not cosmetic. Electrical distribution losses — transformers, PDUs and wiring together — average roughly 10–12% of a data centre's total energy draw, so the equipment chosen to carry that path has a direct efficiency consequence, not just a fit-for-purpose one.

RPP in a data center — when a panel without a transformer is the right link

A remote power panel has no transformer of its own, so it is roughly the footprint of one raised-floor tile. It exists specifically to sit downstream of equipment that has already done the step-down.

Specify an RPP when a floor PDU or stand-alone transformer upstream is already stepping voltage down, and the remaining job is only to split that supply into panelboards near the racks it feeds. An RPP is the wrong choice as a first piece of equipment in a distribution path, because it has nothing to convert — there is no transformer inside it to step anything down.

PDU in the data center — the monitoring spectrum, and what each step buys

Monitoring is not a fixed feature set but a spectrum. At the bottom, a rack PDU is functionally an unmonitored power strip. Moving up the range adds metering of current, voltage, power, power factor, and temperature or humidity sensing. At the top, individual outlets can be switched on or off remotely.

Vendor material sometimes labels this spectrum as four discrete tiers — "basic / metered / switched / per-outlet metered." That exact four-way naming is a common convention, not a formally defined industry taxonomy, so it is used here descriptively rather than as a cited standard. The underlying spectrum itself — from no monitoring to per-outlet remote switching — is well supported.

The top of that spectrum has a real operational payoff. A switched or per-outlet metered PDU is what makes it possible to power down an unloaded or lightly loaded redundant module while keeping the required redundancy elsewhere, which raises the load factor — and therefore the efficiency — of the units that stay in service. Power-factor visibility matters for a related reason: poor power-factor management raises cable losses and can draw grid-operator penalties, which is part of why power factor is one of the values tracked on a metered unit.

Single-phase or three-phase input, and why phase balance matters

A PDU's input can be single-phase or three-phase. Three-phase spreads a larger total load across three legs, which is why it is the default at floor-PDU scale and often at rack-PDU scale too — but only if the load actually lands evenly across the three legs.

Unbalanced loading across the legs of a three-phase system drives higher current between legs and generates waste heat at the supply transformer. Redistributing load to rebalance the phases reduces that waste heat and the cooling load that comes with it. High-efficiency transformers built into the PDU itself run roughly 2–3% more efficient than generic ones — a separate lever from phase balance, but one that compounds with it.

Working the numbers: an unevenly loaded row on a three-phase feed

Take an 18-rack row fed from a 400/230 V three-phase PDU, with six racks wired to each of the three phase legs.

Balanced case. Each rack is budgeted at a steady 4 kW design load: 6 × 4 kW = 24 kW per leg. Current per leg is power divided by voltage: 24,000 W ÷ 230 V = 104.3 A per leg, on all three legs.

Uneven case. Two racks on Phase A are later provisioned at 8 kW each, while the other four on that leg stay at 4 kW: (4 × 4 kW) + (2 × 8 kW) = 16 kW + 16 kW = 32 kW on Phase A. That leg's current becomes 32,000 W ÷ 230 V = 139.1 A. Phases B and C are untouched and stay at 104.3 A.

The row's total load is now 32 + 24 + 24 = 80 kW, for an average of 26.67 kW per leg — 115.9 A, call it 116 A. A feeder or main breaker sized to that row average would already be undersized for Phase A alone, which is drawing 139.1 A. The figure that has to drive the design is the worst-loaded leg, not the total load divided evenly by three. That is why phase balance has to be checked at commissioning and re-checked every time racks are re-provisioned.

Input and output configuration

A PDU's integral transformer commonly steps a North American distribution voltage of 480/277 V down to a utilisation voltage of 240/120 V or 208/120 V for IT equipment. In Europe the utilisation voltage is already 230 V single-phase, so an additional isolation transformer stepping down to 120 V is generally unnecessary and adds only conversion losses. This is a regional difference an enquiry has to state explicitly rather than assume.

On the output side, the C13 and C19 outlet and plug couplers common on rack PDUs are standard connector types defined under IEC 60320, not an interchangeable convenience. Mismatching coupler type against IT equipment cords is a basic but common specification gap.

Breaker coordination inside the unit

A rack PDU sits at the rear of the rack, so its thermal rating has to be judged against the IT equipment's exhaust air temperature rather than its inlet temperature — the two can differ by tens of degrees. That is why new rack PDU products are recommended to be designed for a maximum ambient of at least 60°C.

Breakers inside the unit are themselves temperature-sensitive and not interchangeable on that basis. Magnetic-hydraulic breakers are commonly rated for use up to roughly 85°C, while some less expensive thermal-magnetic breakers are rated only to about 40°C. That mismatch causes nuisance tripping if the breaker rating is not checked against the rack's actual ambient.

Two clauses in the IEC family bound how far a rack PDU's ambient rating can practically go. IEC 61439 sets a 35°C ambient ceiling, with a two-hour excursion allowance to 40°C, and a stricter humidity limit than the equivalent ASHRAE thermal class. IEC 60950-1's 70°C touch-temperature ceiling on metal receptacle and handle surfaces, minus roughly a 10°C allowance for internal conductor temperature rise, is what caps a practical rack PDU ambient rating near 60°C.

Separately, NFPA 70 Article 700.32 requires overcurrent protective devices in emergency power systems to be selectively coordinated with both the supply-side and load-side devices, so that a downstream fault trips only the breaker nearest the fault rather than cascading upstream and dropping unaffected loads.

Dual-feed and automatic transfer

Dual-corded IT equipment is normally fed by two separate rack PDUs, one on each side of the rack, each drawing from a different leg of a diverse upstream source. If one leg drops, the equipment keeps running on the other supply, with no switching action needed.

A rack automatic transfer switch (rack ATS) exists to give single-corded equipment the same protection. It carries two input cords and switches the load to the secondary source within about half an AC cycle if the primary fails — fast enough that the equipment's own internal power-supply capacitance rides through the gap.

How much redundancy the distribution path needs at all is set by the facility's target resiliency rating. ANSI/TIA-942-C defines four rated levels: Rated-1 has a single distribution path with no component redundancy; Rated-2 keeps a single path but adds redundant components; Rated-3 requires at least one active plus one standby (N+1) path and must be concurrently maintainable without disrupting operations; Rated-4 requires at least two simultaneously active (2N/N+N) paths. The rating chosen upstream is what decides whether dual-feed and a rack ATS are needed at all, or optional.

What a data center PDU enquiry must state to get a comparable quotation

"PDU" alone is not a specification. An enquiry needs to state:

1. The tier — floor PDU, RPP, rack PDU or busway tap-off.
2. Input configuration — single- or three-phase, voltage, and step-down requirement if any.
3. Output configuration — outlet or coupler type and count, or panelboard and breaker layout.
4. The monitoring grade needed, stated as function (remote current per outlet, temperature sensing, remote switching) rather than a vendor label.
5. Breaker type and the ambient temperature the unit will actually see at the rear of the rack, not the room's cold-aisle figure.
6. Whether dual-feed and automatic transfer are required, driven by the facility's target resiliency rating.

Leaving any one of these open lets two "equivalent" quotations diverge without either side being wrong about what they priced.

Frequently asked questions

What is the difference between a data center PDU and a remote power panel?

A PDU (floor-standing) has its own step-down transformer and feeds downstream equipment, including RPPs. An RPP has no transformer — it only splits a supply that something upstream already stepped down, in roughly the footprint of one raised-floor tile.

Is a switched or metered PDU worth the extra cost over a basic strip?

It depends on the decision it needs to support. Metered and switched units let you see per-outlet load, catch phase imbalance, and power down unloaded redundant capacity — a basic strip gives none of that visibility.

Why does phase balance matter if the total row load is within capacity?

Because a feeder or breaker has to be sized to the worst-loaded leg, not the row's average. An unevenly loaded leg can exceed its rating even while the row's total draw looks comfortably under budget.

Do I need a rack ATS if my equipment is already dual-corded?

No — dual-corded equipment already draws from two independent rack PDUs and needs no switching. A rack ATS exists specifically to give single-corded equipment that same protection.

What is the single most common gap in a PDU enquiry that stalls a quotation?

Leaving the monitoring grade unstated, or naming a vendor label instead of the function needed (per-outlet current, remote switching). The tier and phase are usually specified; monitoring and rear-of-rack ambient are the two most often left open.