What Is a PDU?
A PDU (power distribution unit) is the device that distributes electrical power from a facility's infrastructure to the individual servers, switches, and storage devices inside a data center rack. It is the final link in the power chain — sitting between the building's electrical panel (or UPS output) and the equipment that actually does the computing.
At the simplest level, a PDU takes a high-amperage power feed and splits it into multiple lower-amperage outlets that devices plug into. But modern data center PDUs do far more than that: they monitor power consumption in real time, allow remote switching of individual outlets, track environmental conditions, and integrate with data center infrastructure management (DCIM) platforms.
PDUs are not optional accessories. Every server rack in every data center uses them. The question is not whether to use a PDU, but which type and how to size it for the workload.
The Four Types of PDUs
PDUs are classified by the level of monitoring and control they provide. Each step up adds capability — and cost.
1. Basic PDU
A basic PDU is essentially a high-grade power strip. It receives a single power input and distributes it to multiple outlets with no monitoring, no switching, and no network connectivity. Basic PDUs are reliable, inexpensive, and appropriate for environments where power monitoring is handled at the panel or circuit level rather than the rack level.
- Monitoring: None
- Remote control: None
- Best for: Labs, development environments, cost-sensitive deployments
- Typical cost: $50 – $300
2. Metered PDU
A metered PDU adds real-time power monitoring. A built-in display (and often a network interface) shows total amperage, voltage, and sometimes per-outlet or per-branch readings. This allows operators to see how much power a rack is drawing without needing to check the electrical panel.
- Monitoring: Aggregate and/or per-outlet amps, volts, watts, kWh
- Remote control: Read-only (no outlet switching)
- Best for: Colocation environments where tenants need to track their own power usage, production racks where capacity planning matters
- Typical cost: $300 – $800
Why metering matters: Without per-rack metering, operators cannot accurately allocate power costs to tenants, identify overloaded circuits before they trip, or plan capacity expansions. Metered PDUs pay for themselves in a few months through better billing accuracy alone.
3. Switched PDU
A switched PDU adds remote on/off control for individual outlets. Through a web interface, SNMP, or CLI, operators can power-cycle a specific server without physically visiting the rack. This is critical for remote management of equipment that has become unresponsive.
- Monitoring: Same as metered
- Remote control: Per-outlet on/off and power cycling
- Best for: Remote or lights-out data centers, managed hosting providers, any environment where physical access is limited
- Typical cost: $800 – $1,500
4. Intelligent (Monitored + Managed) PDU
An intelligent PDU combines metering, switching, environmental monitoring, and full network management into a single device. These units include sensors for temperature, humidity, and airflow; support SNMP traps and RESTful APIs; and integrate with DCIM platforms for automated alerting and capacity management.
- Monitoring: Per-outlet power (amps, volts, watts, kWh, power factor), environmental sensors
- Remote control: Per-outlet switching, outlet groups, scheduled power cycles, firmware updates
- Best for: Enterprise data centers, high-density GPU/AI racks, any environment requiring granular visibility and automation
- Typical cost: $1,500 – $4,000+
Comparing PDU Types
| Feature | Basic | Metered | Switched | Intelligent |
|---|---|---|---|---|
| Power distribution | Yes | Yes | Yes | Yes |
| Aggregate metering | No | Yes | Yes | Yes |
| Per-outlet metering | No | Some | Some | Yes |
| Remote switching | No | No | Yes | Yes |
| Environmental sensors | No | No | No | Yes |
| DCIM integration | No | Limited | Yes | Full |
| Price range | $50–$300 | $300–$800 | $800–$1,500 | $1,500–$4,000+ |
Rack-Mount PDU vs Floor-Standing PDU
PDUs come in two physical form factors that serve different roles in the data center power hierarchy.
Rack-Mount PDUs
Rack-mount PDUs install vertically (0U) along the sides of a server rack or horizontally (1U/2U) within the rack. They are the most common type and are what most people mean when they say "PDU." A typical 42U rack has two vertical rack-mount PDUs — one for the A feed and one for the B feed.
Floor-Standing (Cabinet) PDUs
Floor-standing PDUs, sometimes called RPPs (remote power panels), are larger units that sit on the data center floor. They receive power from the main switchgear or UPS systems and distribute it to multiple rack-mount PDUs via branch circuits. A single floor-standing PDU might feed 20 – 40 racks.
In large data center deployments, the power flows from: utility/generator → ATS → UPS → floor-standing PDU → rack-mount PDU → server.
Single-Phase vs Three-Phase PDUs
The choice between single-phase and three-phase power for your PDUs depends on rack density and total power draw.
Single-Phase PDUs
Single-phase PDUs operate on a single phase of AC power, typically at 120V (North America) or 230V (Europe/Middle East). They are simpler, cheaper, and sufficient for racks drawing under 5 kW.
- Voltage: 120V or 208V (L6-30) in North America; 230V in UAE/Europe
- Max capacity: Typically 2.8 kW (120V/20A) to 8.6 kW (208V/30A)
- Best for: Low-density racks, networking equipment, storage arrays
Three-Phase PDUs
Three-phase PDUs handle higher power loads by distributing the load across three phases. This is essential for high-density racks running GPU servers, ASIC miners, or AI training clusters that draw 15 kW to 50 kW or more per rack.
- Voltage: 208V or 415V three-phase (varies by region)
- Max capacity: 17 kW to 70+ kW depending on amperage and voltage
- Best for: High-density compute, GPU racks, mining hardware, AI workloads
UAE context: In the UAE, data centers typically use 400V/415V three-phase distribution. High-density GPU and mining racks at Rax Data facilities run on three-phase PDUs rated for 30A to 60A per phase, supporting rack densities up to 50 kW per cabinet.
A/B Redundant PDU Feeds
In any production data center, every rack should have two independent PDUs connected to separate power paths. This is called A/B redundancy and is a fundamental requirement of data center power redundancy design.
How A/B feeds work:
- The A-feed PDU connects to Power Path A (one UPS, one set of distribution panels, one utility feed)
- The B-feed PDU connects to Power Path B (a completely separate UPS, panels, and feed)
- Servers with dual power supplies plug one cord into the A PDU and the other into the B PDU
- If the entire A path fails — including the A-feed PDU itself — the B path continues to power every device without interruption
This design aligns with 2N power redundancy configurations where each path is independently capable of carrying the full rack load.
How to Size a PDU for Your Rack
Choosing the right PDU requires matching four specifications to your workload:
Step 1: Calculate Total Rack Power
Add the nameplate power draw (in watts) of every device that will be installed in the rack. Include servers, switches, storage, and any ancillary equipment. For planning purposes, use 80% of nameplate as the expected average draw.
Step 2: Apply the 80% Rule
Never load a PDU above 80% of its rated amperage for continuous operation. This is an NEC (National Electrical Code) requirement for branch circuits and a universal best practice. A 30A PDU should carry no more than 24A continuously.
Step 3: Match Voltage and Phase
Select single-phase for racks under 5 kW and three-phase for anything above. Higher voltage (208V or 415V) delivers more power per amp, reducing copper costs and improving efficiency.
Step 4: Count Outlets
Ensure the PDU has enough outlets (and the right outlet types) for all devices in the rack, plus 20% spare capacity for future expansion. Common outlet types include C13 (10A, for servers), C19 (16A, for high-draw equipment), and C39/Saf-D-Grid for ultra-high-density applications.
| Rack Power | Recommended PDU | Phase | Notes |
|---|---|---|---|
| Under 5 kW | 30A single-phase (208V) | Single | Standard networking/storage racks |
| 5 – 15 kW | 30A three-phase (208V) | Three | Mixed compute racks |
| 15 – 30 kW | 60A three-phase (208V) | Three | High-density GPU racks |
| 30 – 50+ kW | 60A three-phase (415V) | Three | AI training clusters, ASIC miners |
How to Choose the Right PDU
Selecting a PDU is a decision that balances cost, capability, and operational requirements. Here is a practical decision framework:
- Lab or development environment: Basic PDU. Spend on compute, not power management.
- Colocation tenant: Metered PDU minimum. You need to track your own power consumption for billing verification and capacity planning.
- Managed hosting provider: Switched PDU. Remote outlet control is essential when you manage thousands of racks and cannot send a technician for every reboot.
- Enterprise or hyperscale: Intelligent PDU. Per-outlet visibility, DCIM integration, and environmental monitoring are table stakes for environments where a single rack failure can trigger an SLA breach.
- High-density AI/GPU racks: Intelligent three-phase PDU with per-outlet metering. At 30 – 50 kW per rack, granular power monitoring is not optional — it is how you prevent circuit overloads and optimize power distribution across phases.
For operators deploying GPU clusters or mining hardware in the UAE, Rax Data facilities provide three-phase power distribution with metered PDUs as standard, ensuring accurate power tracking and redundant A/B feeds across all rack positions. Contact our team to discuss power requirements for your deployment.
Frequently Asked Questions
What is a PDU?
A PDU (power distribution unit) is a device that distributes electrical power to servers, networking gear, and other equipment inside a data center rack. PDUs receive power from the facility's electrical panel or UPS system and split it into multiple outlets that individual devices plug into. They range from simple power strips to intelligent units with per-outlet monitoring, switching, and remote management.
What is a PDU in a data center?
In a data center, a PDU is the final link in the power chain between the building's electrical infrastructure and the IT equipment in each rack. Data center PDUs are rated for higher amperage (typically 30A to 60A per unit) than consumer power strips and are designed for continuous 24/7 operation. Most data center racks use two PDUs (A and B feeds) connected to separate power paths for redundancy.
What are the four types of PDUs?
The four main types are: (1) Basic PDUs, which simply distribute power with no monitoring; (2) Metered PDUs, which display real-time amperage and voltage readings; (3) Switched PDUs, which allow remote on/off control of individual outlets; and (4) Intelligent (or monitored) PDUs, which combine metering, switching, per-outlet power data, environmental sensors, and network-based remote management.
How do you size a PDU for a server rack?
Add up the maximum power draw of all devices in the rack (in watts or amps), then select a PDU rated for at least 20% more than that total. The 80% rule means you should never load a PDU above 80% of its rated capacity for continuous operation. For a 20 kW rack, you would need PDUs rated for at least 25 kVA. Also match the PDU's voltage (120V, 208V, or 400V) and phase (single or three-phase) to your facility's power distribution.
What is PDU power redundancy (A/B feeds)?
PDU power redundancy uses two independent PDUs (called A and B feeds) in every rack, each connected to a separate power path from the facility's electrical system. Servers with dual power supplies plug one cord into the A-feed PDU and the other into the B-feed PDU. If either PDU or its upstream power path fails, the other continues to power the equipment without interruption.
Planning Your Data Center Power Infrastructure?
Rax Data & Energy provides three-phase power distribution with metered PDUs, A/B redundant feeds, and infrastructure engineered for high-density compute.
Contact Us Our Infrastructure