For the hypothetical 1 MW IT scenario in this article, N requires four 0.5 MVA blocks, N+1 installs five, and 2N installs eight. Those counts follow from the same load assumptions. They do not prove that an actual facility can survive a failure. N+1 adds one equal block to the duty count; 2N duplicates the entire duty set. The difference becomes clearer when you calculate the duty requirement before choosing the arrangement.
Define the boundary and assumptions
This is a facility-level capacity sandbox, not a UPS sizing schedule or a single-line design. Assume 1 MW of IT input, planning PUE 1.25, facility power factor 0.95, a user-selected 20% planning reserve and equal blocks of 0.5 MVA. These figures are illustrative, not prescribed minimums or standard equipment selections.
| Input | Example value | Meaning here |
|---|---|---|
| IT load | 1 MW | Modeled electrical input to IT |
| Planning PUE | 1.25 | Facility-overhead scenario multiplier |
| Power factor | 0.95 | Assumed value at the modeled facility boundary |
| Reserve | 20% | Explicit allowance above the operating apparent load |
| Block capacity | 0.5 MVA | Equal hypothetical units for counting |
A block could represent an early planning abstraction; it is not automatically a transformer, generator or UPS module. In particular, not every facility load necessarily belongs on the IT UPS. Schneider Electric’s discussion of UPS deployment distinguishes IT and cooling backup considerations. Establish each equipment boundary before translating this example into a project schedule.
Calculate N before adding redundancy
- Facility demand scenario: 1 × 1.25 = 1.25 MW.
- Operating apparent load: 1.25 ÷ 0.95 = approximately 1.316 MVA.
- Planning duty requirement: 1.316 × 1.20 = approximately 1.579 MVA, calculated without intermediate rounding.
- Duty blocks: round 1.579 ÷ 0.5 upward to the next whole number, giving four.
Three blocks provide only 1.5 MVA, below the 1.579 MVA planning requirement. Four provide 2.0 MVA. The approximately 0.421 MVA between that installed duty capacity and the requirement comes from block rounding. It is distinct from the 20% reserve already included in the requirement.
Compare the three arrangements
| Arrangement in this model | Installed blocks | Total capacity | Count after the stated loss |
|---|---|---|---|
| N: one duty set | 4 | 2.0 MVA | 3 after losing one block: 1.5 MVA |
| N+1: duty set plus one block | 5 | 2.5 MVA | 4 after losing one block: 2.0 MVA |
| 2N: two complete duty sets | 8 | 4.0 MVA | 4 after losing one complete set: 2.0 MVA |
The table compares capacity remaining, not simulated electrical behavior. For N+1, the remaining four blocks cover the selected planning requirement after one block is unavailable. For 2N, each four-block set can cover that requirement. Whether the surviving capacity can actually reach the load depends on connections, protection, controls and the operating state.
There is an important nuance in the N row. Its remaining 1.5 MVA is below the reserved planning envelope but above the 1.316 MVA operating apparent-load assumption. It would be wrong to conclude from these numbers alone that the current load must be dropped. The arithmetic demonstrates loss of the full planning allowance, not a time-domain failure prediction.
Reserve and redundancy are not interchangeable
Reserve changes how much duty capacity the model requires. Redundancy changes how many blocks or sets are installed around that duty requirement. Increasing reserve may push the duty count across an integer threshold; adding one spare block does not increase the declared IT load. Track both choices so that a growth allowance is not added to server demand and then added again as reserve.
At the same operating assumptions, installing 4.0 MVA in the 2N case does not turn the 1.25 MW facility scenario into 4 MW of consumption. Installed apparent capacity, operating real power and annual energy are different quantities. Nor does eight blocks versus five establish a complete construction-cost ratio: distribution, controls, space and other project requirements remain outside this model.
The architecture questions the table cannot answer
Schneider Electric’s UPS configuration explanation describes system-plus-system arrangements in terms of power paths, not merely extra equipment. Treat the following as questions for a design review, not a commissioning procedure:
- Which shared bus, upstream source, control system or room could affect both sets?
- Can the intended load remain supplied during the specified maintenance state?
- Do server power-supply connections support the assumed surviving paths?
- Are the stated block ratings valid for the environment and duty?
- What happens to cooling and control power during the same event?
Hardware requirements matter at the load end too. NVIDIA’s DGX H100 electrical guidance discusses particular energized-PSU and distribution-path requirements. A redundancy label by itself is not enough to establish compatibility. No Tier classification, uptime percentage or fault-tolerance certification is assigned by this article.
Reproduce the example and change one assumption
In the data center capacity planner, enter IT load 1 MW, PUE 1.25, PF 0.95, reserve 20% and block size 0.5 MVA. Compare N, N+1 and 2N without changing the load. Then vary block size to see the effect of rounding. If your starting point is a server list, first calculate its IT load with the AI power calculator; carry IT MW forward, not facility MW.
Sources & further reading
- Schneider Electric: Design Considerations for the Deployment of UPS Systems in Data Centers ↗
- Schneider Electric: UPS Deployment Design Choices for High Availability Applications ↗
- NVIDIA: Electrical Specifications — DGX H100 ↗