Colin Rees, associate director at IES, a global climate tech firm specialising in building-physics software, performance digital twins and consultancy.
The growth of digital infrastructure is bringing a wider energy management problem into focus. As AI and digital services become increasingly important to economic growth, the data centre that support them are under pressure to perform efficiently and reliably, and in many cases, that performance is falling short.
Across the built environment, many assets have high levels of energy wastage, relying on legacy controls, static models and systems that only show part of the picture. They can show what a building is using, but not always why, how energy usage is likely to change over time, and how performance can be optimised.
As demand rises and grid capacity tightens, those shortcomings become more pertinent. The growth of digital infrastructure and subsequent heightened demand on the grid is inevitable, but this doesn’t need to mean that energy management systems underperform.
Data centres are exposing the performance gap at scale
Global data centre electricity consumption is projected to double by 2030, driven largely by AI and high-density computing. In the UK, data centres already account for about 2.5% of electricity use, with this expected to rise fourfold by 2030.
Because data centres run around the clock, even small inefficiencies can quickly become serious cost, carbon and energy resilience issues. Cooling is a major part of that challenge, with traditional cooling systems accounting for up to 40% of a data centre’s total energy demand. Water use is also rising as many operators look for new ways to manage growing heat loads.
But one of the core problems is that poor performance is often built in from the start.
Across the built environment, many design and retrofit decisions are still based on simplified calculations or static models that do not reflect how buildings actually operate. They can miss seasonal change, part-load performance and shifts in occupancy or use. In practice, that can mean oversized plant, inefficient controls, missed hotspots and systems that look fine on paper but fall short in operation. Too often, systems are designed around peak or notional conditions, rather than the other 99.9% of the year when assets actually operate.
This is what we call the performance gap: the difference between how a building is designed to perform and how it performs once it is up and running. In data centres, this can show up as wasted cooling energy, unnecessary water use or infrastructure that struggles to adapt as workloads increase. In other buildings, it can mean heating and cooling systems working against each other, half-empty spaces being fully conditioned, and facilities teams spending more time firefighting than planning ahead.
The building types may differ, but the issue is the same. Too often, complex assets are still being managed with methods that are too simplistic for the demands placed on them.
Reducing energy use at scale using dynamic simulation
Data centres show how the right approach can significantly reduce energy use at scale.
Dynamic, whole-system simulation enables teams to model how a building or facility will perform over time, using real weather data, building fabric, controls, internal loads and different operating scenarios. Put simply, it gives a more realistic picture of how an asset is likely to behave before major decisions are made.
That matters because it lets teams test options before build or upgrade work begins, rather than finding problems only once they are costly and disruptive to fix. It also helps them plan for future conditions, whether that means hotter summers, changing usage patterns, rising IT loads or tighter grid constraints.
One IES project for a hyperscale facility in North America shows the value of this approach. Using dynamic simulation to assess cooling options for a site in a cool, dry climate, the team found that direct evaporative cooling was the best fit, delivering a projected Power Usage Effectiveness (PUE) of 1.16 and a 95% reduction in water use compared with water-cooled alternatives. PUE measures how efficiently a data centre uses energy by comparing the total power used by the facility with the power used by its IT equipment. The closer the figure is to 1.0, the less energy is being lost on cooling and other supporting systems.
While there isn’t one universal solution for every project, whole-system modelling makes it possible to identify the right options for each specific building, climate and operating profile.
The wider lesson
Energy managers need to understand how buildings perform in practice, how systems work together and which interventions will deliver the biggest gains in energy, carbon and cost. Rising demand doesn’t equate to rising waste, but avoiding that outcome requires a move beyond outdated assumptions and towards a more realistic view of building performance.
This article appeared in the June 2026 issue of Energy Manager magazine. Subscribe here.
