Data center power distribution, from the utility feed to the rack
Every watt a server uses has travelled through a chain of transformers, switches, batteries and distribution units. That chain decides how reliable a data centre is, how efficient it is and how quickly it can grow.
The power path in one diagram
The simplified one-line diagram below shows a single path. Real facilities duplicate most of it.
Utility feed or on-site plant (medium voltage)
|
Substation / step-down transformer
|
Main switchgear <---- Standby or prime generators
| (via transfer switch)
UPS (rectifier - batteries - inverter)
|
Power distribution units (PDUs) / remote power panels
|
Rack PDUs
|
Server power suppliesStage by stage
1. Incoming supply
Most facilities take a medium-voltage feed from the utility. A growing number also generate on site. When the on-site plant is the main source, the site is running behind the meter.
2. Transformers and switchgear
Transformers step the incoming voltage down for building distribution. Switchgear provides protection and switching, and it is where the generators connect. A transfer switch moves the load between utility and generator.
3. UPS
The UPS sits between the switchgear and the IT load. In an online, double-conversion UPS, the batteries are always connected to the inverter, so no transfer switch is needed in the IT path. The UPS carries the load for the seconds or minutes it takes generators to start. See data center backup power for how the layers work together.
4. PDUs and remote power panels
Downstream of the UPS, power distribution units split the supply into branch circuits. Wikipedia defines a PDU as a device with multiple outputs designed to distribute electric power to racks of computers and networking equipment, and notes that most large data centers utilize PDUs with three-phase power input and one-phase power output.
5. Rack PDUs
Inside each cabinet, rack PDUs feed the servers. Basic rack PDUs simply distribute power. Intelligent PDUs add metering at the inlet, outlet or branch circuit, remote on and off control and alarms, and some support temperature and humidity sensors.
Redundancy: A/B feeds and Tier levels
To prevent single points of failure, all elements of the electrical system, including backup systems, are typically duplicated. Critical servers are connected to both the A-side and B-side power feeds, an arrangement often used to achieve N+1 redundancy.
Uptime Institute's Tier classification sets how far this goes: Tier I requires UPS and generator; Tier II adds redundant capacity components; Tier III is concurrently maintainable with redundant distribution paths; Tier IV is fault tolerant with independent, physically isolated systems.
Efficiency: losses and PUE
Every conversion stage loses some energy as heat. The standard yardstick is power usage effectiveness. PUE is the ratio of total facility energy to IT equipment energy. An ideal PUE is 1.0. The metric was developed by The Green Grid and published as ISO/IEC 30134-2:2016.
Distribution design affects PUE directly. Fewer conversion stages mean fewer losses.
Designing for growth
Demand from AI workloads is rising quickly, and electrical capacity, not floor space, is often what limits a site. For the scale of demand involved, see data center power requirements. Three questions shape the design:
- Where does the power come from? Grid capacity is often the bottleneck, which is pushing developers to generate on site.
- How much redundancy is needed? The Tier target sets the number of UPS modules, generators and distribution paths.
- What rating will the generators run on? If backup plant may run as primary power, size it on its prime or continuous rating. See prime vs standby power.
Power at the top of the chain
The distribution chain can only be as firm as its source. BridgePower Nuclear, a joint venture of USP&E and BAM Energy, is focused on that first box in the diagram. USP&E supplies thermal power on gas, diesel, HFO or dual-fuel, grid-tied or behind the meter. It brings 24+ years in power, 150+ power projects in 45+ countries, 2,000+ MW in stock and 400+ engineers and project managers.
If you are planning a site and the supply is the constraint, talk to our team.
Frequently asked questions
How is power distributed in a data center?
Power arrives from the utility or an on-site plant at medium voltage, is stepped down by transformers, passes through switchgear and transfer switches to the UPS, then through power distribution units (PDUs) and rack PDUs to the servers. Generators connect upstream of the UPS.
What is a PDU in a data center?
A power distribution unit is a device with multiple outputs that distributes power to racks of servers and network equipment. Data centre PDUs commonly take a three-phase input and provide single-phase outputs, and intelligent PDUs add metering and remote control.
What are A and B power feeds?
Two independent power paths to the same equipment. Critical servers are connected to both, so a failure or maintenance on one path does not drop the load.
What is power usage effectiveness (PUE)?
PUE is the ratio of total facility energy to the energy used by IT equipment. An ideal PUE is 1.0. The metric was developed by The Green Grid and published as ISO/IEC 30134-2:2016.
Does on-site generation change the distribution design?
It moves the main source on site, but the downstream chain of switchgear, UPS and PDUs stays much the same. The main changes are at the point of supply: protection, synchronisation with any grid connection, and redundancy of the generating units.