Beyond PUE & WUE: Advanced Metrics for Measuring Efficiency in AI Data Centers

AI growth is driving demand for high‑performance data centers. This whitepaper explores how rising energy and water use along with the shift to liquid cooling present the need to reassess whether PUE and WUE remain the best efficiency metrics relative to compute power.
April 7, 2026

As AI usage increases, the demand for high-performance data centers has surged. These data centers are the backbone of AI applications, providing the necessary computational power to process vast amounts of data. With this increased power consumption comes the need for greater efficiency in terms of both energy and water usage. The two traditional metrics used to measure this efficiency are Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE). However, with the increased adoption of liquid cooling for high-density chips, exploration of whether PUE and WUE are the best metrics for measuring efficiency in AI data center applications is required, especially when compared to compute power.