Why the data centre conversation now starts with substations

A few years ago, if a data centre client asked about power, the conversation was usually about how many megawatts and how many transformers. Substations sat in the background - important, but treated as a specialist item to be resolved once the campus was designed. Today they are one of the first things clients want to discuss.

That is because the role of the substation has changed. As AI continues to reshape our industry, power has become one of the defining factors in whether a data centre can be delivered at all. Clients are not simply asking whether there is enough capacity on the grid. They are asking how power will reach the site, how quickly it can be connected, how resilient it will be and whether the infrastructure can support future expansion. Those questions all start with the substation.

What fascinates me is not simply that substations have become more important. It is that they have become a critical part of data centre design. For a long time, substation design and data centre design were treated as adjacent disciplines rather than integrated ones. The data centre team owned the design down to the medium voltage switchgear; the high voltage substation was typically resolved with the utility and specialists as a separate workstream. That worked because campus loads were smaller, the utility interface was more predictable, and the substation footprint had modest influence on the wider master plan.

Today, that separation is disappearing. The amount of power required by AI workloads has fundamentally changed the conversation. Decisions about grid connections, utility interfaces, resilience, future capacity and expansion all influence the wider design of the campus. They are no longer engineering details to resolve later. They shape decisions from the very beginning.

Hyperscalers have had substation delivery teams for years. What has changed is the weight those teams now carry in project decisions. Substation strategy increasingly precedes campus strategy - some hyperscalers now integrate substation planning into land acquisition itself, not just engineering delivery. Increasingly, clients are asking consultants not simply whether they can design a data centre, but whether they understand the electrical infrastructure that underpins it.

I think that is a significant shift. It tells us the industry no longer sees substations as supporting infrastructure but as a strategic part of the development itself. Naturally, that also changes when we need to start thinking about them.

One of the biggest challenges I see is treating power as something to resolve once the land has been secured. In reality, power feasibility - grid availability, connection timelines, capacity for future expansion, utility willingness to support the load - should be tested before a site is purchased, let alone before a data centre is designed. A site that looks perfect on every other criterion can be undeliverable if the grid cannot support the load in the timeframe the client needs.

Can the local grid support the proposed development? How long will a connection take? Will upgrades be needed? Should the client consider local generation, battery storage or another solution?

Those questions do not just influence electrical design. They influence programme, investment decisions and, in some cases, whether a project proceeds at all. AI workloads have made this more acute. They draw significantly higher power densities and, until recently, created sudden load steps that stressed the grid in ways older workloads did not. Hardware manufacturers are beginning to address this through rack-level capacitance and on-board power smoothing. But the underlying density challenge remains, and utilities are still catching up with the pace of change. That is why we are seeing power due diligence become an increasingly important part of early project development, and why utilities themselves are becoming more cautious about how these loads are connected to their networks.

Another consequence of this shift is in how our teams work together. The industry does not need to reinvent the engineer - there is already a talent shortage in both data centre design and substation engineering. What we need are engineers on both sides who understand each other’s disciplines well enough to design together rather than in parallel. A decision made within the substation affects the layout, phasing and long-term resilience of the entire campus; decisions made by the data centre team directly shape the electrical infrastructure required to support them. Bridging that gap does not mean replacing specialists. It means bringing specialist knowledge together much earlier in the process.

Because successful projects are no longer delivered by independent workstreams that eventually meet in the middle. They are delivered by integrated teams that understand the relationship between power infrastructure and data centre design from day one. As AI continues to drive larger campuses, higher power densities and more ambitious digital infrastructure, I believe that integration will become one of the defining characteristics of successful projects.

The conversation may begin with substations. But what clients are really asking runs deeper: whether their consultants can design the whole system as one integrated whole, not as separate workstreams that eventually meet.