This piece answers a question the spec sheets hide: how much power can one outlet of your PDU actually deliver? The answer is not the PDU's total rating, and it is not the branch breaker — it is whatever the smallest component in the chain allows, and in a standard rack that component is a $1 connector. Every rating below is taken from the IEC 60320 standard as documented by connector-industry guides, and from Vertiv's official PowerIT rack PDU installer/user guide; each figure is cross-checked across at least two independent sources, and where sources disagree the piece drops the number instead of averaging it.

What are the three limits stacked in a rack PDU?

LayerGoverning ruleLimit (US home, 120 V)What it caps
Branch circuitNEC 210.20(A): breaker ≥ noncontinuous + 125% × continuous1,920 W continuous (20 A / 120 V)The sum of everything plugged in, for the whole strip
PDU nameplateManufacturer's input ratingWhatever the unit lists (often ≥ the branch, sometimes lower on metered units)The unit's total throughput
Per-outlet connectorIEC 60320 ratings; UL/CSA per-outlet deratingC13: 10 A IEC / 12 A UL ≈ 1,200–1,440 W; C19: 16 A ≈ 1,920 W; NEMA 5-15R: 12 A per Vertiv; 5-20R: 16 A per VertivThe single device in that outlet

The connector ratings are the layer people skip, so pin them down. IEC 60320-1 rates the C13/C14 pair at 10 A, 250 V — and this figure is stated identically by Vertiv's official rPDU manual, by a PDU-industry selection guide, and by a connector-manufacturing sourcing guide. North American catalogs often list C13 higher (one guide says "UL ratings permit up to 15 A"); Vertiv's own table lands at 12 A UL/CSA for C13 outlets on its PDUs — which is the 80% rule applied to the connector, the same 125%-derated arithmetic that gives the branch its 1,920 W. Where sources disagreed — one connector guide's temperature table put the C19 mating-face limit at 155 °C where the PDU-industry guide puts C19 at 70 °C (155 °C belongs to the C21/C22 class) — we dropped the disputed number entirely rather than pick a side.

How do connector amps turn into watts?

Multiply by voltage (ours — this is arithmetic on the verified amp ratings, not a new spec): at US household 120 V, a C13 outlet's ceiling is 10 A × 120 V = 1,200 W at the IEC rating, or 12 A × 120 V = 1,440 W at the UL/CSA per-outlet rating Vertiv uses. A C19 outlet at 16 A is 1,920 W — notably, exactly the continuous budget of an entire 20 A branch. At 240 V, the same connectors double: a C13 carries up to 2,400 W (IEC 10 A) and a C19 up to 4,000 W, which is why every high-density rack design — and our own branch-circuit piece — keeps arriving at the same conclusion: voltage is the cheapest amp you will ever buy.

Now run the numbers on real hardware (ours, using the ratings above). A dual-PSU server drawing 1,900 W is fine on C13 in a US rack: each power supply pulls ~950 W through its own cord, about 7.9 A — comfortably inside even the IEC 10 A limit, and the redundancy gives you a second failure domain for free. The same 1,900 W through a single-PSU box on one C13 cord at 120 V is 15.8 A — above the 12 A UL per-outlet rating and the IEC 10 A rating both. That single fact explains why high-draw servers ship with C20 inlets: the C19/C20 pair exists precisely because the C13's 10 A ceiling ran out somewhere in the blade-and-big-router era. Vertiv's own manual matches the split: C13 outlets for standard 1U/2U gear, C19 outlets for the high-draw tier, NEMA L6-20R (250 V, 16 A) listed for 240 V feeds.

The aggregate view is where home labs get caught. A 24-outlet C13 strip can theoretically carry 24 × 1,440 W = 34.6 kW of connected devices — but it is fed by one 20 A branch at 1,920 W continuous. The breaker governs the sum; the connector governs each socket; and the ratio between the two is not a contradiction, it is the design. A PDU is not 24 little circuits. It is one circuit with 24 delivery points, each of which must stay under its connector cap while the fleet stays under the breaker.

What does an overloaded connector actually look like?

Not like a tripped breaker — that is the whole point. The C13's limiting component is not the copper, it is the plastic and the contact spring. A connector-industry sourcing guide puts the C13 mating-face temperature limit at 70 °C per IEC 60320-1 and describes the failure sequence precisely: a cord that carries the current fine at room temperature "can soften, discolour and lose contact pressure in a hot enclosure," and the result arrives "months later as intermittent reboots rather than as a clean open circuit." A rack running 24/7 (continuous load, by definition) with warm intake air is exactly that enclosure. The 70 °C limit is a property of the plastic and spring, "not of the copper" — the wire inside the cord is sized for the current, but the mating interface is not sized for the heat that a loose, oxidized or overloaded interface generates on its own.

For hot-running gear, the standard already drew the escalation path: the C15/C16 pair is the same 10 A rating with a 120 °C mating-face limit, used for equipment that radiates heat into its own inlet — dense PoE switches, blade-class boxes. Both a PDU selection guide and a connector sourcing guide list the C15 at 120 °C against the C13's 70 °C (three independent sources agree on the 120 °C figure). If a device's inlet runs hot, the fix is not a "better" C13 cord — the fix is the connector class designed for the temperature. (One sourcing note from the same guides: the C-family is keyed so hot-class and cold-class connectors do not intermate freely; treat the keying as the spec, not an inconvenience.)

What is the actual checklist?

Three tallies, in order (ours): First, the biggest single device against the connector. Anything drawing more than ~1,440 W at 120 V does not belong on a C13 outlet, period — it needs a C19 outlet, a 240 V feed, or two cords on two PDUs. A big line-interactive UPS recharging after a Deep outage is the classic surprise load: check its charge-current spec, not just its output rating. Second, the sum against the breaker. 1,920 W continuous on a 20 A / 120 V branch, per the NEC piece — and remember the branch carries the PDU's own overhead plus anything else on the circuit. Third, the 80% rule at every layer. Vertiv's own per-outlet tables (12 A on a 15 A-class receptacle, 16 A on a 20 A-class) are the same 125% continuous-load discipline showing up one more time; a device running at 100% of any nameplate in a warm rack is a device with no margin. Dual-PSU servers get one upgrade for free: put the two cords on two different PDUs — ideally two branches or two phases — and a tripped breaker stops meaning a down server.

And the cheapest diagnostic in the whole chain is a hand. A C13 plug face that is warm-to-hot at idle, or a cord that has taken a slight set near the mold, is the failure mode announcing itself months before the intermittent reboots start. The breaker will never tell you; the connector will, if you check.

What are the caveats?

Voltage matters more than the connector in 240 V regions. The 1,440 W figure is a US-120 V number; on European 230 V mains a C13 at 10 A is 2,300 W and the per-outlet limit rarely binds first — this piece's arithmetic is US-home-specific and says so. Per-outlet derating is manufacturer-specific. Vertiv lists C13 at 12 A UL/CSA; one industry guide mentions UL ratings up to 15 A for the bare connector; the honest number for your rack is whatever your PDU's manual states per outlet — read the table, not the marketing. One source conflict was dropped. The two guides disagree on the C19 mating-face temperature class (70 °C vs 155 °C); since 155 °C is the C21/C22 class in both guides' own C13/C15/C19/C21 comparison tables, we exclude the disputed figure rather than publish one side. The failure-mode description comes from a connector-industry guide, not from an insurance investigation — it is consistent with the 70 °C standard and with ordinary contact physics, but it is an industry account, not a lab study. And nothing here is a substitute for your local electrical code — the NEC citations in our branch-circuit piece are US-specific, and so is this one.