By Keiran O’Gorman, Technical Lead, Major Projects, Eland Cables
Five years ago, a developer signing off a datacenter masterplan could bereasonably confident that future demand would follow familiar patterns.Capacity requirements might grow faster or slower than expected, but thedirection of travel was generally understood.
AI has changed that equation. Across the industry, operators are discovering that assumptions made only a few years ago around power density, expansion timelines and future customer requirements are being revisited much sooner than expected. Facilities designed with significant headroom are being reassessed, while projects still in development are being asked to accommodate workloads that barely featured in planning discussions when the designs first emerged.
Few would have predicted how quickly demand would change.
As organisations race to deploy AI technologies, much of the conversation naturally focuses on processors, servers and cooling systems. Yet one of the biggest questions facing the sector sits further down the chain. Can the electrical infrastructure supporting these facilities adapt quickly enough to keep pace?
When forecasts start to drift from reality
Higher rack densities, greater power requirements and the pace of change incomputing infrastructure are forcing operators to revisit plans sooner than anticipated. In some cases, facilities that were considered comfortably future-proof when they were designed are now being evaluated against requirements that exceed original expectations.
This does not mean existing facilities are incapable of supporting AI workloads.In many cases, it simply means operators are reassessing infrastructure that was designed around a different set of assumptions.
Power is becoming the defining constraint
For years, power was largely discussed in terms of availability.
Could a site secure sufficient grid capacity? How long would a connection take? What level of resilience was required?
Those questions remain important, but AI is broadening the discussion.
Attention is shifting beyond how much power can be secured and towards how effectively that power can be distributed, managed and expanded over the lifetime of a facility.
This is bringing electrical infrastructure into sharper focus. Medium-voltage cable systems may not attract the same attention as AI hardware, but they determine how power moves through a site and how easily a facility can adapt when requirements change.
Once major electrical infrastructure is installed, modifying it can become complex, disruptive and expensive, particularly within operational environments where uptime remains the overriding priority. That means decisions made during the earliest design stages can influence a facility’s adaptability many years later
The bottlenecks are not always where people expect
Grid constraints receive much of the attention, and rightly so. Across many regions, operators are already competing for finite power availability.
Yet some of the most immediate challenges emerge within the datacenter itself. Space for additional cable routes. Capacity within existing distribution systems. Access to infrastructure corridors. Provision for future expansion.
They often become critical later, when operators need to support higher-density deployments or accommodate customer requirements that were never envisaged in the original design brief. And by that point, the cost and complexity of making changes can rise considerably.
What appears to be a relatively modest design decision at the beginning of aproject can have a significant impact on how easily that facility adapts severalyears later.
Building flexibility into the foundation
Perhaps the biggest lesson from the AI boom is that certainty has become harderto achieve. Few operators would claim to have a clear picture of what AI infrastructure requirements will look like five years from now.
The most resilient projects are often those that acknowledge that reality fromthe outset. Rather than attempting to predict exactly what future demand willlook like, they create room for infrastructure to evolve alongside it.
That might mean preserving space for additional circuits, allowing greater capacity within distribution networks, or considering how infrastructure can be expanded without major disruption.
Their importance often becomes fully apparent when requirements change andoperators need room to adapt.
Planning for a future that keeps moving
The datacenter sector has always adapted to new technologies. AI is unlikely tobe the last development that challenges established assumptions.
What makes the current situation unusual is the pace of change.
Infrastructure plans that looked robust only a few years ago are already beingre visited. Expansion strategies are being accelerated. Design assumptions are being tested earlier than expected.
For operators and developers, that creates an important consideration: how much of tomorrow’s flexibility is being determined by today’s decisions?
Electrical infrastructure may not be the most visible part of a modern datacenter, but it is becoming one of the most consequential. The ability to accommodate future demand, respond to changing customer requirements and support long-term growth will depend heavily on choices made long before additional capacity is needed.
In an industry shaped by constant technological advancement, adaptability is becoming one of the most valuable forms of capacity a facility can possess.
