Power first: how electrical infrastructure is shaping data centre delivery

In data centre construction right now, one question tends to come up before anything else: how much power can actually be secured, and when? It’s no longer a detail to resolve once a scheme is defined. In many cases, it’s the factor that determines whether that scheme takes shape at all.

The pace of demand, driven largely by AI, has pushed power infrastructure well beyond what many networks were designed to support. Connection timelines are stretching, capacity is uneven, and assumptions that would have felt conservative even a few years ago are being tested much earlier. Developers aren’t just responding to demand anymore, they’re working within the limits of what the grid can realistically deliver.

That constraint is starting to show up in the shape of projects, with schemes increasingly being phased or reconfigured to reflect when power becomes available.

It also brings electrical infrastructure into sharper focus much earlier in the process.

Medium voltage networks have always formed the backbone of a data centre’s power distribution. What’s changing is the point at which they begin to influence decision-making. Cable routes, distribution strategy and capacity allowances are no longer technical details to be worked through once layouts are fixed. They need to be understood and coordinated alongside them, because they directly affect how infrastructure can be delivered and how much flexibility remains later.

On constrained sites especially, that influence is tangible. Routing corridors compete for space with drainage, structural elements and access routes. Decisions about segregation, protection and future expansion all have spatial consequences. Once those routes are committed, the ability to adapt becomes limited, not just from an electrical perspective but across the wider scheme.

That matters because data centre requirements aren’t holding still.

Power densities continue to climb, particularly with the growth of AI and high-performance computing. Clients are refining their requirements later in the process, sometimes in response to their own end users, sometimes because technology itself is evolving. What looked like sufficient capacity at concept stage can start to feel restrictive by the time construction is underway.

The gap between those early assumptions and eventual demand is where pressure starts to build.

In practice, that tends to show up in one of two ways. Either infrastructure is pushed harder than originally intended, with less headroom than designers would ideally prefer, or projects have to find ways to overlay additional capacity, often at greater cost and complexity.

Neither outcome is particularly efficient, and both reinforce the same point. Decisions that sit in the background at design stage can have very visible consequences once a facility is operational.

Cables sit right at the centre of that.

They don’t attract the same attention as switchgear or major plant, but they define how power is physically distributed across a site. Once installed, cabling is difficult to modify without disruption, particularly in live environments. Capacity, routing and installation conditions therefore carry long-term implications that are not always obvious at the outset.

Those implications become more pronounced as utilisation increases. Cables are now expected to carry higher loads, more consistently, and often in more demanding environments. The margin between theoretical performance and real-world conditions is narrower, which means that factors such as soil thermal resistivity, ambient temperature and installation method start to carry more weight. This is where the interaction between design and delivery becomes more critical.

A cable system might meet the required specification on paper, but if installation conditions differ from what was assumed, performance over time can be affected. That might relate to congestion in service corridors, changes in routing, or simply the practical realities of working on a busy construction site.

At the same time, programme pressure is compressing the window in which these issues can be worked through.

Longer lead times for key electrical equipment mean that elements of the power system are often on the critical path from an early stage. That drives earlier commitment on infrastructure and reduces the opportunity to refine design as more information becomes available. It keeps projects moving, but it also places greater emphasis on getting those early decisions right.

For contractors, it changes the sequencing challenge. Electrical infrastructure is no longer something that can follow behind civil works in a predictable way. It needs to be coordinated alongside them, with less room for adjustment once construction is underway.

All of this contributes to a delivery model that feels less linear than it once did.

Data centre projects have always involved coordination across multiple disciplines, but the level of interdependency is increasing. Power availability influences design. Design influences routing and installation. Installation conditions influence long-term performance. Each step feeds back into the others.

That’s particularly evident when requirements change mid-project, which is becoming more common. The ability to respond quickly depends on how much flexibility has been built into the system at an earlier stage. Where that flexibility is there, adaptation is manageable. Where it isn’t, change tends to bring delay or additional cost.

This is why relatively modest interventions are taking on greater importance. Allowing space for additional circuits, maintaining access to key routes, and selecting cable specifications that can accommodate higher loads are not headline design moves. They don’t fundamentally change how a scheme looks on paper, but they make a significant difference to how it performs when conditions shift.

In a more stable environment, those decisions might be seen as conservative. In the current one, they are increasingly pragmatic.

What sits behind all of this is a broader change in the role of power within data centre development.

It’s no longer just an input that needs to be delivered efficiently once everything else is decided. It’s a constraint that shapes what is possible, and when. It influences not just the technical design, but the commercial and programme decisions that sit around it.

For developers and contractors, that alters the nature of the challenge.

Delivering quickly still matters. So does cost. But there is a growing emphasis on whether a scheme can adapt to changing requirements without losing momentum. That depends less on any single component, and more on how well infrastructure choices reflect the reality that demand is unlikely to stand still.

In that sense, power isn’t just supporting data centre construction anymore. It’s defining the boundaries within which that construction takes place.

Keiran O’Gorman is technical lead major projects at Eland Cables