In short. A 2N architecture provides two complete and independent power paths, A and B, each able to carry the whole critical load on its own, from generation through switchgear and distribution to the equipment. N+N is the same idea stated for a plant where N is more than one unit. The word that matters is independent: two paths that share a fuel system, a bus, a controller or a cable trench are not 2N, however many generators stand in the yard.
From extra capacity to a second path
An N+1 plant adds a spare unit inside one system. A 2N plant adds a second system. The distinction is where the redundancy boundary sits. In N+1, the generators are redundant but the bus, the transfer switch and the distribution usually are not. In 2N, every element between the two generator plants and the load exists twice, so that a bus fault, a switchboard fire, a controller failure or a maintenance shutdown on the A side leaves the B side untouched. The critical load connects to both, either through dual power supplies in the equipment or through a static transfer switch where the equipment has a single input. The family of options, including N+2 and distributed redundancy, is compared under N, N+1, N+2, 2N and distributed redundant systems.
What each path must contain
| Layer | Path A | Path B | Independence test |
|---|---|---|---|
| Generation | Generator plant A, sized for full critical load | Generator plant B, sized for full critical load | Separate fuel storage, transfer, batteries and controls |
| Transfer | ATS or utility-generator switchgear A | ATS or switchgear B | No shared bus, no common coupler left closed |
| UPS | UPS system A | UPS system B | Separate bypass sources, separate batteries |
| Distribution | Switchboards and PDUs A | Switchboards and PDUs B | Separate rooms or fire compartments, separate cable routes |
| Load connection | Cord A to PSU A | Cord B to PSU B | Each PSU alone can carry the device |
The table is also the audit. For every row, the question is whether a single event can remove both columns at once. The A/B power distribution page follows the two paths through a building in detail.
The shared elements that quietly break 2N
On paper, 2N is easy to draw. On site, the two paths tend to converge in places nobody drew.
- Fuel. Two generator plants fed from one bulk tank through one transfer pump have one fuel system, and the day it fails both plants stop within hours.
- Controls. A single master controller or a single SCADA server commanding both plants can take both down with one software fault or one wrong command.
- Routes. A and B feeders run through the same trench or the same riser share every excavation, flood and fire in that route.
- Rooms. Two switchboards in one room share its fire, its cooling failure and its roof leak.
- Bypass. UPS A and UPS B that share one maintenance bypass source converge whenever either is on bypass.
- Ties. A bus coupler between A and B, provided for flexibility, becomes a common point the moment someone leaves it closed.
These are common-cause failures, and they are the main reason a measured 2N system performs worse than its model. The analysis method is described under common-mode and common-cause failure engineering.
The load end: dual cord and single cord
A 2N infrastructure delivers its value only if the equipment can use both paths. Servers, storage and network switches with two power supplies draw from A and B simultaneously and continue on either alone. Each supply must be rated for the full device load, and both cords must genuinely terminate on different paths, which is easy to get wrong in a crowded rack. Single-cord equipment, including many chillers, pumps and older medical devices, cannot connect to two paths directly; a static transfer switch or a rack-level transfer switch selects between A and B in a few milliseconds, and that switch then becomes part of the reliability analysis as a shared component. Even dual-cord equipment can contain a single internal element after the two supplies, so the facility can provide 2N inputs but cannot guarantee redundancy inside a device beyond what its manufacturer designed. See dual-cord and single-cord critical loads.
Loading rules in a 2N system
In normal operation each path carries half the critical load, so that when one path fails the other picks up the whole of it. That constraint is the one most often violated as a facility grows.
Load per path (normal) ≤ 50 percent of path capacity
Path capacity is the lowest-rated element in that path, which may be the UPS, a switchboard or the generator plant after derating rather than the largest number on the drawing. Worked example (illustrative): a data hall with a 1,200 kW critical load has two paths each with a 1,500 kW UPS and a 2,000 kVA generator plant derated to 1,450 kW. Each path normally carries 600 kW, 40 percent of its capacity, and on loss of path A the B path carries 1,200 kW, 80 percent, within limits. If the hall grows to 1,600 kW, each path carries 800 kW normally, but on a failure the survivor must carry 1,600 kW against a 1,450 kW generator plant, and the architecture has silently become N. Capacity management against this rule is a DCIM function, as discussed under data centre power.
Variants: 2(N+1), distributed redundant and when 2N is justified
Where even a single unit failure within one path during an outage on the other is unacceptable, each path is built as N+1 in its own right, giving 2(N+1). Where cost is the constraint, a distributed redundant arrangement spreads the load across three or more systems, each carrying two-thirds of its rating, so that any one can fail; it saves capacity against 2N at the price of more complex transfer logic. The decision between them follows the consequence and duration of an outage, not the prestige of the label. Hyperscale and colocation data halls, trading floors and some pharmaceutical processes justify 2N; most industrial plants, offices and hospitals outside the theatre block do not, and are better served by a well-built N+1 with concurrent maintainability, which is covered under concurrent maintainability engineering.
Common mistakes
- Calling two generator plants on one bus 2N. The bus is the common point and a bus fault or a maintenance shutdown removes both plants.
- Feeding both generator plants from one bulk fuel tank and pump. The fuel system is the real redundancy boundary and it is N.
- Letting normal load per path drift above 50 percent as the facility grows. The surviving path overloads at the first failure and the architecture is 2N in name only.
- Plugging both cords of a server into outlets on the same PDU. The rack is single-path regardless of what the building provides.
- Leaving the A to B tie closed for convenience. A fault on either side propagates to both.
What to send us
Send the single-line diagram showing both paths from generators to the load, the critical load with its growth forecast, the UPS and generator ratings on each path, a site plan showing rooms and cable routes for A and B, and the fuel system layout. Tell us whether the loads are dual-cord, single-cord or mixed. Use the application engineering enquiry page.
Questions engineers ask
What is the difference between 2N and N+N?
They describe the same principle. 2N is used when N is a single unit, for example two 1,000 kVA generators each able to carry the full load on its own path. N+N is used when N is itself a group, for example two plants of three 500 kVA sets each. In both cases each side must independently support the full critical load to the defined boundary.
Can the two paths share a utility supply?
They often do, because a second DISCOM feeder is not always available, and the utility is then a shared element upstream of the redundancy boundary. The 2N claim applies from the generator plants downward, and the design accepts that a utility failure is a normal event handled by both plants starting. Where two independent utility feeders exist, the boundary moves upstream.
Does 2N mean I can skip maintenance planning?
No. With one path shut down for maintenance the facility is at N, and a failure on the live path loses the load. 2N makes maintenance possible without an outage, which N+1 behind a single switchboard does not, but the exposure during each maintenance window must still be scheduled, kept short and accepted deliberately.
Is 2N worth it for a hospital?
For the whole hospital, rarely. For operating theatres, intensive care and imaging, a dual-path arrangement at the final distribution is common and is often required by healthcare standards, while the wider building runs on N+1 generators with a single transfer system. Matching the architecture to each load class costs far less than applying 2N everywhere.
Technical content reviewed by Srinivas Reddy, Managing Partner, Reddy Generators. Last updated 19 October 2026.