PUE
Power Usage Effectiveness: the ratio of total data centre energy to the energy used by IT equipment; 1.0 is perfect efficiency.
Last refreshed: 6 May 2026 · Appears in 1 active topic
Is a 1.1 PUE actually sustainable when the facility drawing it pulls 500 MW?
Timeline for PUE
Mentioned in: Where the next data centres should go
Data Centres: Boom and Backlash- What is a good PUE for a data centre?
- An industry average PUE is about 1.58 (Uptime Institute 2024). hyperscale facilities typically achieve 1.1–1.2. A PUE below 1.5 is generally considered efficient; below 1.2 is considered leading-edge.Source: Uptime Institute
- Why are AI data centres harder to keep efficient?
- GPU server racks for AI training generate significantly more heat per unit of floor space than traditional servers, requiring more intensive cooling. Without purpose-built liquid cooling, AI density can push PUE above a facility's designed ceiling.Source: Uptime Institute
- What does PUE stand for and why does it matter?
- PUE stands for Power Usage Effectiveness — the ratio of total facility power to the power used by IT equipment. A PUE of 1.0 would be perfect; the industry average is about 1.58. It matters because it directly measures how much electricity is wasted on cooling, lighting, and infrastructure rather than computing.Source: Uptime Institute
- How is liquid cooling changing data centre PUE?
- Liquid cooling (direct-to-chip and immersion) is now standard in AI-optimised facilities because GPU rack densities generate too much heat for air cooling to handle efficiently. Liquid cooling can keep PUE at 1.1-1.2 even at very high densities, whereas air-cooled AI facilities can see PUE above 1.5.Source: Uptime Institute
Background
Power Usage Effectiveness (PUE) is the standard metric for measuring data centre energy efficiency. Calculated as total facility power divided by IT equipment power, a PUE of 1.0 would mean all electricity goes to servers with no overhead; the global industry average is approximately 1.58 according to Uptime Institute's annual benchmarking. hyperscale facilities typically achieve PUEs of 1.1 to 1.2 through purpose-built cooling and advanced heat management.
AI training workloads are creating pressure on PUE in two directions. High-density GPU server racks generate substantially more heat per unit of floor space than traditional servers, requiring more aggressive cooling that can raise PUE. Simultaneously, some hyperscalers are achieving very low PUEs in custom AI campuses by integrating liquid cooling directly at the chip level. Uptime Institute's 2024-2025 research found AI infrastructure is widening the efficiency gap between leading-edge and average facilities.
PUE is widely reported in operator sustainability disclosures but critics note it measures only relative efficiency, not absolute energy consumption. A 1.1 PUE at a 1 GW facility consumes FAR more energy than a 1.5 PUE at a 10 MW facility. Regulators in Ireland and Spain have begun incorporating PUE thresholds into data centre planning conditions.