Power Infrastructure Basics
Data center power infrastructure converts utility power (typically medium voltage) to the clean, conditioned power that IT equipment requires. The path from utility entrance to server power supply passes through: utility transformer, main switchgear, UPS system, power distribution units (PDUs), and rack-level PDUs.
Utility Transformer
Steps down medium voltage (13.8 kV typical) to low voltage (480V or 208V) for distribution within the facility.
Main Switchgear
Distributes power from the transformer to UPS systems and other loads. Provides overcurrent protection and isolation capability.
UPS System
Provides clean, conditioned power and battery backup. Protects against power quality events (sags, surges, harmonics) and brief outages.
Power Distribution Units (PDUs)
Distributes power from UPS to rack-level PDUs. Provides metering and monitoring for power consumption by zone.
Rack PDUs
Distributes power to individual servers within a rack. Smart PDUs provide per-outlet metering and remote switching.
UPS Systems
UPS (Uninterruptible Power Supply) systems protect IT equipment from power quality events and provide battery backup during brief utility outages. Double-conversion UPS provides the cleanest power: it continuously converts AC to DC and back to AC, isolating the load from all power quality events. Line-interactive UPS is less expensive but provides less protection.
UPS runtime is not the primary purpose
Generator Systems
Generator systems provide power during extended utility outages. Diesel generators are the most common for data centers, they start within 10–15 seconds and can run for days with adequate fuel supply. Natural gas generators eliminate fuel storage requirements but depend on gas supply availability during emergencies.
Generator sizing is critical: generators must be sized for the full facility load plus startup surge, not just the average load. Undersized generators that cannot carry the full load during a utility outage are the most common power infrastructure failure mode in enterprise data centers.
Cooling Infrastructure Basics
Every watt of power consumed by IT equipment is converted to heat that must be removed from the data center. Cooling infrastructure removes this heat: maintaining safe operating temperatures for all equipment. The cooling system must be sized for the maximum power load, not the average load.
CRAC/CRAH Units
Up to 5 kW/rackComputer Room Air Conditioning/Handling units. Standard for traditional data centers. Inadequate for high-density deployments.
In-Row Cooling
5–15 kW/rackCooling units placed between server rows. More efficient than perimeter cooling for medium-density deployments.
Rear-Door Heat Exchangers
10–30 kW/rackAttaches to rack rear, captures heat at source. Requires chilled water. Effective for high-density compute.
Direct-to-Chip Liquid
20–60 kW/rackDelivers coolant directly to CPU/GPU heat spreaders. Required for sustained GPU operation at full TDP.
PUE: Measuring Energy Efficiency
PUE (Power Usage Effectiveness) measures how efficiently a data center uses energy. PUE = Total Facility Power / IT Equipment Power. A PUE of 1.0 is perfect: all power goes to IT equipment. A PUE of 2.0 means the facility uses as much power for cooling, lighting, and other overhead as it uses for IT equipment.
1.0–1.2
Excellent
Hyperscale data centers
1.2–1.5
Good
Modern enterprise facilities
1.5–2.0
Average
Typical enterprise data centers
2.0+
Poor
Legacy facilities
AI Workload Power Density
AI workloads have fundamentally changed the power density requirements of enterprise data centers. A single NVIDIA H100 GPU has a 700W TDP. An 8-GPU server (DGX H100) consumes approximately 10.2 kW, 2–3x the power of an entire rack of traditional servers. A rack of 4 DGX H100 servers requires 40+ kW, far beyond the capacity of air cooling.
Most facilities cannot support AI workloads without upgrades