In a data centre, backup power can be built as one centralised UPS room feeding the whole hall, or as distributed units placed at the rack row or even inside the cabinet. Centralised design wins on serviceability and battery management at scale, while distributed design cuts distribution loss and limits the blast radius of a failure; the right choice depends on hall size, rack density and how the operator wants to contain fault domains.
A centralised system puts one or a few large UPS modules in a dedicated power room, along with the bulk battery bank. A bus, usually at 415 V three-phase, runs to the IT halls and steps down at the rack row. The protected load is fed from a single engineered system with a single battery enclosure that a technician can walk up to. This is the classic model for large enterprise and colocation halls.
The advantage is concentration. Batteries live in a conditioned room with proper ventilation and fire detection, monitoring is unified, and a failed module is swapped by a technician who is already on site. For a hall drawing several megawatts, one well-run centralised plant is simpler to own than dozens of scattered ones.
Distributed UPS moves the power protection close to the load. At row level, a module sits in or above the rack row and feeds just that row. At rack level, a unit mounts in the cabinet and protects a single rack. The IT load is fed at or near its final voltage, so there is little or no long low-voltage run between the UPS and the servers.
This is attractive at high density, where pushing forty to one hundred kilowatts per rack over a long bus means thick copper and real conductor loss. Putting the UPS at the row shortens that run to a metre or two. Our data-centre power configurations are outlined at https://www.upsboss.com/data-center/.
The decisive difference is often the copper. A centralised UPS feeds the hall over a long low-voltage bus carrying high current, and conductor loss rises with the square of that current. At AI-class densities those losses become a meaningful slice of facility power and a real cooling load. Distributed units feed short final runs, so that loss largely disappears.
The trade is that you buy more, smaller UPS modules and spread the batteries around, which costs more per kilowatt than one large plant. The question is whether the saved distribution loss and floor space are worth the premium, and at high density they usually are. The current product range is at https://www.upsboss.com/products/.
Centralised systems concentrate risk. A fault in the power room, a battery event or a maintenance error can take the whole hall, which is why they are built 2N for critical loads. Distributed systems spread that risk: a failed row-level module blacks out one row, not the building, and the failure is contained by design.
That containment has a real operational value. During a distributed-unit fault, the rest of the hall keeps running and the affected row can be serviced without touching shared infrastructure. The blast radius is the single strongest argument for distributed architecture in large, mixed-tenancy halls.
Centralised wins on day-to-day care. Batteries in one room are easier to inspect, temperature-control and replace, and monitoring is a single system rather than fifty. Distributed lithium batteries are smaller and often module-swappable, but they are spread across the floor, which complicates a unified view of state of health.
For an operator with a strong facilities team, centralised is comfortable. For an operator who wants to avoid ever entering a battery room during a fault, distributed is cleaner. Neither is wrong; they suit different staffing and risk postures.
Centralised batteries need a room with floor loading and ventilation sized for the bank, which is space taken away from revenue-generating IT. Distributed lithium is lighter and mounts close to the load, returning that floor to cabinets. In a dense hall where every square metre is billable, giving floor back to IT can outweigh the higher per-kilowatt UPS cost.
The calculation is local: weigh the value of reclaimed IT floor against the premium of many small units and the cost of managing batteries spread across the hall. There is no universal answer, only a building-specific one.
Send us the hall layout, density and tenancy model and our engineers will compare centralised and distributed options with a loss and footprint estimate. Review the configurations at https://www.upsboss.com/data-center/ and the units at https://www.upsboss.com/products/.
Key takeaway: centralised UPS simplifies service and battery management at scale, while distributed UPS cuts distribution loss and contains faults to a row. Choose by density, tenancy and how much IT floor the battery room would cost you.
Is distributed UPS always better for new data centres?
No. It leads at high density and mixed tenancy, but a low-density single-tenant hall with a good facilities team is often cheaper and simpler to run as one centralised plant.
Does distributed design remove the need for a generator?
It does not. Both architectures still feed from the same utility and generator; distribution only changes where the UPS sits relative to the racks.
Which is easier to keep under warranty?
Centralised, because batteries and modules are in one accessible room. Distributed spreads assets across the floor, which needs disciplined tracking to avoid lapsed maintenance on a forgotten unit.
Contact: Frank Zhang
Phone: +86-135 5688 8641
Email: frank@upsboss.com
Add: Jufeng Road, Guangming Street, Guangming District, Shenzhen City, Guangdong Province, China
