What Is Data Center Power?

Data center power refers to the electrical systems, architectures, and management practices that supply, condition, and distribute electrical energy to computing infrastructure within a data center facility. It encompasses the path from utility service entrance through transformers, switchgear, uninterruptible power supplies, and power distribution units to individual servers and network equipment, as well as the parallel systems for backup generation and energy recovery. Power is one of the two dominant operating costs in a data center (alongside cooling), and its reliability directly determines whether the facility can meet its contractual service availability commitments.

Data centers are among the largest single consumers of electrical power in any urban or industrial district. Hyperscale facilities operated by cloud providers can draw 100 megawatts or more, requiring dedicated utility substations and transmission-level connections. Smaller enterprise data centers typically consume between 1 and 20 megawatts. The engineering of reliable, efficient power delivery at this scale draws from power systems engineering, electrical distribution design, and industrial controls.

Power Delivery and Distribution

The electrical distribution path in a data center follows several conversion and conditioning stages. Utility alternating current (AC) power enters at high voltage, is stepped down through transformers, and feeds automatic transfer switches that can rapidly switch between utility and generator sources during an outage. Uninterruptible power supplies (UPS) systems bridge the gap between a utility interruption and generator startup, providing conditioned power for seconds to minutes from battery banks or flywheel energy storage. Power distribution units (PDUs) then branch the conditioned power to individual equipment racks, often at 208V or 415V three-phase. Some facilities are moving toward 380V high-voltage direct current (HVDC) distribution, which eliminates multiple AC-to-DC conversion steps and can reduce distribution losses by several percentage points. Design standards from ASHRAE 90.4, which establishes energy performance baselines for data center electrical systems, define how mechanical load component (MLC) and electrical load component (ELC) targets are expressed and measured.

Energy Efficiency Metrics

Power usage effectiveness (PUE) is the primary metric for measuring data center energy efficiency. It is defined as the ratio of total facility power consumed to the power delivered specifically to IT equipment: a PUE of 1.0 would represent a perfectly efficient facility with no overhead from cooling, lighting, or distribution losses. Industry average PUE has declined from approximately 2.0 in the early 2000s to around 1.55 for typical enterprise facilities, with hyperscale operators such as Google reporting fleet-wide PUE values near 1.09. The Green Grid developed PUE in 2006, and it has since been incorporated into ISO/IEC 30134-2:2016 as an international standard. Complementary metrics include water usage effectiveness (WUE) and carbon usage effectiveness (CUE), which account for cooling water consumption and emissions associated with the facility's power source. IBM's data center overview describes how operators use PUE and related metrics to benchmark efficiency against industry peers and track progress against sustainability goals.

Backup Power and Resilience

Continuity of power requires redundant generation capacity. Diesel generators, sized to carry the full facility load, are the standard backup source, with startup times of 10 to 30 seconds bridged by UPS batteries. Generator fuel supply must support continuous operation for a minimum of 12 to 72 hours, depending on the facility's tier classification and contractual commitments. Higher-tier facilities deploy N+1 or 2N generator redundancy so that any single generator can fail without affecting output. Newer installations are incorporating natural gas generators, fuel cells, and on-site renewable generation to reduce emissions and dependence on diesel supply chains. AWS's documentation on data center availability infrastructure describes how distributed availability zones allow traffic to shift away from a facility experiencing a power event, reducing the criticality of local backup capacity.

Applications

Data center power engineering applies across a range of facility types and deployment scenarios, including:

• Hyperscale cloud facilities requiring utility-scale power procurement and distribution
• Colocation data centers with diverse tenant power requirements and metering
• High-performance computing and AI training clusters with GPU-dense racks exceeding 40 kW each
• Edge data centers operating from limited utility feeds in remote or constrained locations
• Mission-critical financial and healthcare facilities requiring the highest tier of power resilience