After topology, the next decision on an uninterruptible power supply is whether the rating arrives in one block or in modules. That choice sets your redundancy method, your service model and your first-day capital cost, and it is difficult to reverse once the room is laid out. Both architectures are current across the UPS systems catalog, so this is a live specification decision rather than a generational one.

This guide is one of four spokes beneath the data center UPS and energy storage hub, which sets out how topology, storage, sizing and frame architecture fit together. Read it alongside how to size a UPS and UPS topologies explained.

Key Takeaways

  • Redundancy cost is set by module size rather than frame size: a 500 kW load takes eleven 50 kW modules for N+1 and pays 10% for redundancy, or three 250 kW modules and pays 50%.
  • Modular frames let you buy the ceiling and populate for day one, which is the argument whenever the contracted load is above the installed load.
  • Monolithic units are usually cheaper per kW at full build, carry fewer connectors and control paths, and put a physical wall between redundant halves.
  • Hybrid designs with large power cores buy online serviceability without a 50 kW module connector count, and give up fine-grained redundancy in exchange.
  • Decide the redundancy scheme before the frame. A 2N decision doubles the electrical room; a modular N+1 decision does not.

What modular actually means#

A modular UPS is a frame that accepts power modules of a fixed rating. Capacity is added by inserting modules, and redundancy is achieved by installing one more module than the load requires. On decentralized designs each module carries its own rectifier, inverter, controls and static bypass, so no shared component determines whether the frame survives a module fault.

The ABB DPA 250 S4 is a clear example: 50 kW modules, a 300 kW installed frame built from six of them, and a 250 kW N+1 rating. The ABB DPA UPScale ST 80 applies the same idea at 10 kW and 20 kW module granularity in an 80 kW frame. At the megawatt end, the ABB MegaFlex DPA uses 250 kW slide-in modules and extends to 2,000 kW.

Module size is the specification that matters#

Redundancy cost follows module size, and frame size has almost nothing to do with it.

11 modules of 50 kW carry a 500 kW load at N+1 — ten modules meet the load, the eleventh is the redundancy, and it costs 10% of the load.

3 modules of 250 kW carry that same 500 kW load — two meet the load, the third is the redundancy, and it costs 50%.

That arithmetic explains the shape of the current product range. The Schneider Electric Galaxy VL covers 200 to 500 kW from 50 kW modules, and the Schneider Electric Galaxy VS covers 10 to 150 kW for smaller rooms. The Eaton 93PM G2 uses 50 kW power modules and offers internal N+1 up to 200 kW in its North American form.

Table 1 — Module granularity across the modular range#

Model Module size Frame range Redundancy step
ABB DPA UPScale ST 80 10 kW / 20 kW to 80 kW One module
Eaton 93PM G2 50 kW Internal N+1 to 200 kW (North America) One module
Schneider Electric Galaxy VL 50 kW 200–500 kW One module
ABB DPA 250 S4 50 kW 300 kW installed, 250 kW N+1 One module
ABB MegaFlex DPA 250 kW to 2,000 kW One module

Read the redundancy step column against the frame range and the trade is visible. A one-module step is cheap at 50 kW granularity and expensive at 250 kW.

Where monolithic still wins#

A monolithic UPS delivers its rating from one set of power electronics. It is usually cheaper per kW when the load is known and stable, it has fewer connectors and control paths, and it is straightforward to parallel for redundancy at system level.

Table 2 — Monolithic and hybrid frames in the catalog#

Model Architecture Rating range Parallel ceiling
Eaton Power Xpert 9395P Monolithic 200–1,200 kVA across 15 models
Vertiv Liebert EXL S1 Monolithic 250–1,200 kW in 12 steps 9.6 MW
Riello Multi Sentry MST Monolithic 160 and 200 kVA current
Vertiv Liebert Trinergy Cube Hybrid, 400 kW cores 800 kW–1.6 MW per unit 12.8 MW
Vertiv Liebert APM2 Modular, SiC hybrid 10–600 kW

Redundancy in the monolithic world is N+1 at unit level: two 800 kW units carrying a 700 kW load. That buys more redundant capacity than a modular frame would, and it also puts a physical wall between the two halves of the system.

The hybrid case#

Some machines sit between the two. The Vertiv Liebert Trinergy Cube is built from hot-serviceable 400 kW power cores, scales from 800 kW to 1.6 MW in one unit and parallels to 12.8 MW. The Vertiv Liebert APM2 covers 10 to 600 kW in modular form with a silicon carbide hybrid converter.

Large power cores give you online serviceability without the connector count of a 50 kW module design. They do not give you fine-grained redundancy, so the real question is which failure and service behavior you want to buy.

Six questions before the frame is fixed#

  1. What is the day-one load, and what is the contracted ceiling? A frame sized for the ceiling and populated for day one is the argument for modular.
  2. What is the module size, and what does N+1 cost at that granularity?
  3. Who swaps modules, and does the site accept online swapping under load at all?
  4. Is the load likely to move? Modules can follow a phased build in a way a monolithic unit cannot.
  5. What is the part-load efficiency curve? A lightly populated frame and a lightly loaded monolithic unit behave differently, and ENERGY STAR publishes tested efficiency curves across the load band, not a single point.
  6. What does the electrical room allow? Clearances, working space and overcurrent protection follow NFPA 70, and frame depth often decides the layout.

Question two is where the capital comparison lives. Current purchase benchmarks by rating are in the data center UPS pricing guide, and the redundancy step should be priced against those figures before the architecture is chosen.

Standards and reliability context#

Both architectures are specified and tested under IEC 62040-3, so the classification code is comparable across them. Thermal and airflow planning should follow the ASHRAE Datacom series, because modular frames concentrate heat differently from a single large cabinet.

On the reliability side, the Uptime Institute annual outage analysis continues to report power as a recurring cause of significant outages, and human error during maintenance is a repeated theme. Online module swapping reduces the number of full-system transfers a technician has to perform, which is an operational argument as much as a specification one.

Grid-side conditions are moving too. The NERC reliability assessments document tightening reserve margins in several regions, and the US Department of Energy tracks data center load growth against that backdrop. Where thin margins push a site toward more on-site generation, generator lead times and wider backup power lead times become part of the same decision.

FAQ: modular and monolithic UPS architecture#

What is a modular UPS?#

A frame that accepts power modules of a fixed rating, where capacity is added by inserting modules and redundancy is achieved by installing one more than the load requires. On decentralized designs each module carries its own rectifier, inverter, controls and static bypass, so a module fault does not depend on a shared component to be survivable.

Is modular or monolithic cheaper?#

Monolithic is usually cheaper per kW at full build. Modular is usually cheaper across a phased build, because the frame is bought once and populated as load arrives, and because redundancy costs one module where a monolithic pair costs a whole unit. The comparison turns on how far day-one load sits below the contracted ceiling.

What is N+1 redundancy in a UPS?#

N is the capacity that carries the load. N+1 adds one redundant unit or module on top of it. On a modular frame that means one extra module: the ABB DPA 250 S4 installs six 50 kW modules for a 250 kW N+1 rating. On a monolithic pair it means a whole second unit, each sized to carry the load alone.

Can you add modules to a UPS later?#

Within the frame's rating, yes — that is the point of the architecture. The limit is the frame, so the ceiling has to be specified at purchase even when only part of it is populated. Adding capacity beyond the frame rating means a second frame.

Can modules be swapped while the load is live?#

On hot-swappable designs, yes, and it is one of the stronger operational arguments for modular: an online swap avoids a full-system transfer. Whether the site permits it is a separate question, and some operators require a transfer regardless of what the machine supports.

Next step#

Compare module sizes, frame ratings and published efficiency side by side in the UPS systems hub, where every model page lists its architecture, rating steps and current production status.