Nearly one in four data centre operators is failing to track the power consumption of its primary sites, according to a late-2025 study by 451 Research, part of S&P Global Market Intelligence, commissioned by Janitza.
The findings, based on a survey of 208 data centre professionals, suggest that while the industry is racing to build capacity for AI, parts of the sector are still missing some of the most basic visibility into how efficiently their facilities are running.
Using Power Usage Effectiveness (PUE) as a key benchmark, the research found that just over half of respondents reported a PUE of between 1.5 and 2.0. However, 23% said they were not tracking the metric at all.
That is a striking figure at a time when power availability is becoming one of the biggest constraints on data centre growth. As operators look to support more AI workloads, the pressure is not just on securing enough power, but on understanding exactly how it is being used across a site.
Data centres are also facing greater scrutiny from both members of the public, as well as politicians. In Edinburgh, the sector is facing a moratorium over building new data centres due to concerns surrounding power and water usage, with councillors on Edinburgh Council questioning the ‘green’ credentials of proposed data centres in the area.
Risk to AI growth
While public perception is one thing, the report from 451 Research also argues that the lack of energy usage tracking is quickly becoming a business risk rather than just an operational blind spot. Without accurate, real-time energy data, it suggests that operators may struggle to scale AI-ready infrastructure safely, plan capacity effectively, or protect equipment from increasingly volatile electrical loads.
AI is only making that challenge harder. According to the study, highly dynamic AI workloads can drive power fluctuations of between 40% and 70% within milliseconds, creating additional strain on electrical and cooling systems and increasing the risk of outages or equipment damage.
The report notes, “In an environment where milliseconds matter, flexibility and data expertise are the critical differentiators.”
That points to a wider shift taking place across the market. Monitoring is no longer just about producing a headline PUE number for sustainability or efficiency reporting. Increasingly, operators are being pushed towards far more detailed visibility across the whole power chain, from grid connection through to individual racks.
According to the research, that level of monitoring is becoming more important as rack power densities rise towards 40-120 kW and AI models continue to grow in size and complexity. It suggests that operators with better access to high-resolution energy and power-quality data are more likely to identify faults early, support predictive maintenance, extend the life of critical equipment, and avoid unplanned downtime.
In that sense, the study’s findings go beyond a warning about poor monitoring habits. They underline how quickly power management is becoming a competitive issue for data centre operators, particularly those hoping to monetise AI capacity without exposing themselves to greater operational risk.
