Trial / Demo
Demand for digital services keeps rising, with AI workloads and population growth adding pressure on data centre capacity. To meet this demand, more sites are being built, and electricity use continues to grow.
Faced with these challenges, designers, operators and owners need clear indicators on site energy use, and the ability to evidence the impact of energy efficiency measures and where opportunities for future improvements lie. This article explains why static Power Usage Effectiveness (PUE) calculations at design stage are insufficient, and how dynamic simulation provides a more realistic picture of energy performance and cost.
PUE is defined in the ISO/IEC 30134-2:2026 Information technology — Data centres key performance indicators standard as the ratio between total facility energy and IT energy over the same period.
PUE = EDC / EIT
EIT includes servers, storage, network devices, and the stations used to operate and monitor the site.
EDC includes EIT plus all supporting systems such as UPS units, switchgear, PDUs, chillers, cooling towers, pumps, CRAHs, CRACs, DX units, lighting, security, and fire protection.
In the standard, PUE is a measured value which uses real energy readings taken over a full year of operation to report how the facility truly performs. It is not based on design values or a single snapshot.
However, design stage predictions typically report a peak PUE. While this is a useful indicator for capacity checks, peak PUE is calculated at the highest expected IT load and at the hottest design-day conditions, so is not representative of the expected operation. Designers use extreme outdoor dry bulb and wet bulb temperature design set-points to size chillers, pumps, CRAHs, electrical systems, and distribution equipment - typically within a spreadsheet.
The resulting estimates of cooling and electrical losses at this single operating condition provide the peak PUE, which is used to:
However, peak PUE also has clear drawbacks. This approach:
A data centre design can achieve a good peak PUE on paper and yet still perform poorly for many hours of the year.
Real operation evolves hour to hour, with IT loads fluctuating, outdoor temperature and humidity shifting through the day and across the seasons, and chillers, fans, and pumps running at part load while control logic responds to alarms, setpoints, and maintenance activity.
A static spreadsheet cannot reflect these conditions. It acutely cannot show how chilled water temperature resets influence the PUE in shoulder seasons, how CRAH or fan wall control affects fan energy, how humidity control shapes cooling performance in a hot and humid climate, how efficiencies change at part load, or how free cooling varies with real weather.
IES address these analysis gaps by modelling for dynamic PUE - an annualised PUE profile simulated from hourly climate data, hourly IT loads, HVAC system components, and the plant control logic - enhancing the resolution on cooling and electrical load needs. Dynamic PUE reveals short term behaviour, such as cooling swings during humid periods, efficiency shifts at different IT loads, changes caused by chilled water temperature reset or CRAH control, and spikes linked to humidity, or free cooling. Dynamic PUE condenses these hourly effects into an annualised figure that reflects seasonal cooling demand, real IT load patterns, part load performance, and the usable free cooling hours. This more accurate PUE provides a far more reliable template for energy cost and emissions planning than a single peak estimate.
The Uptime Institute 2024 Global Data Center Survey reports that average annual PUE has stayed close to 1.56 in recent years. After large improvements between 2007 and 2014, the global average has settled in a narrow band between about 1.55 and 1.59. The flat trend, however, hides strong differences between sites where older facilities push the average higher, while new builds and large modern facilities achieve lower PUE.
Consider a large data centre designed in a hot and humid climate. The facility covers about 30,000m2 split across two main floors and a partial mezzanine. The layout includes data halls, LV rooms, battery rooms, IDF rooms, corridors, storage areas, and a small administrative section with offices and restrooms. The HVAC setup uses DOAS with VRF for non-critical areas and CRAHs with fan walls arranged in a cold aisle and hot aisle configuration for the IT spaces. Cooling is provided by air cooled chillers grouped into three plants and linked by a primary only chilled water loop delivering free cooling when the climate allows. The system operates with a high chilled water supply and return temperature to improve efficiency and extend the free cooling hours.
For this example, the initial analysis reports a peak PUE of 1.53, which appears favourable when compared with common industry figures as shown above.
However, an IESVE model of the same design applies hourly weather data and realistic IT load profiles across the full year, producing hourly energy use for IT equipment, cooling systems, fans, pumps, and other loads, along with a dynamic PUE profile and a yearly PUE value.
In this case, the analysis reports a dynamic PUE of 1.34, which is considerably improved upon the static peak value of 1.56. The dynamic simulation model shows high efficiency during cooler periods when free cooling and higher plant performance are available. While the PUE profile rises, as expected, during the hottest and most humid months, the overall annual impact remains lower than the static peak figure. The dynamic results also highlight periods where control settings can be refined, such as fan control in shoulder seasons or chiller staging at part load. Overall, the yearly performance and operating cost appear more favourable than the conservative static estimate reports. The chart below visualises the PUE profile through the year.
The heat map below helps visualise how cooler months sit mostly in the green and light-yellow range, while the summer period drives PUE higher, with July and August aligning to high intensity in orange and red. The highlighted point marks the peak PUE, making the limitations of static PUE obvious. This figure reflects only a brief moment in the year. Whereas 99.9% of hours fall below this extreme condition, and the surrounding pattern shows how PUE continually changes with climate, IT load variation, and part-load operation.
From a design standpoint, a dynamic model offers clear advantages. This includes being able to test different chilled water temperatures and the impact to PUE across the seasons, compare containment strategies and CRAH control settings on an hourly basis, quantify the benefit of free cooling and both air cooled and water cooled economizer operation in the local climate, and assess how future IT load growth influences PUE and plant loading.
The same model also supports operations by showing when the site is most sensitive to humidity and outdoor conditions, identifying periods where plant behaviour drifts from expected ranges, and helping refine control sequences so that on site adjustments are reduced.
Peak PUE still supports understanding on the power and cooling capacity the site needs and helps avoid under-sizing during extreme weather. On its own, though, a peak or static design PUE does not show how the facility will perform in operation.
A design can look strong at peak yet still consume excess energy for many hours each year. Dynamic simulation closes this gap by generating the dynamic annualised PUE that reflects real climate conditions, actual load patterns, and true system behaviour. Teams have access to more accurate data for their design decisions, supporting more accurate energy and cost forecasts, sustainability goals and clearer direction on where improvements matter most.
Dynamic PUE modelling shifts the industry from designing for rare peak conditions to understanding how a data centre performs during the reamaining 99.9 percent of the year. By capturing hourly weather, real load behaviour, and true system interactions, we can replace conservative estimates with evidence-based performance insight. Providing designers, owners, and operators a clearer view of where energy is actually consumed, how efficiency changes over time, and which adjustments yield meaningful improvements. As digital demand grows and efficiency targets tighten, dynamic simulation provides the level of accuracy and accountability needed to build and operate data centres that are both resilient and genuinely energy-efficient.
Check out our data centre services or join our upcoming webinar on Thursday 26th February 2026 to see first-hand how dynamic simulation and a whole facility modelling approach can help data centre teams design, optimise and retrofit their facilities with confidence.
