Executive Summary
- UK facilities risk becoming stranded assets when they sit on ‘massive power contracts’ but cannot physically run, due to outdated hardware.
- The commercial value of a data centre means something different since the explosion of AI – and with it – changing performance and cooling demands.
- Here we expose the ‘Megawatt Myth’ and ask in an industry obsessed with size, is it just a big illusion?
Let’s be honest, the data centre industry has always had a subtle obsession with size. Executive boardrooms and marketing campaigns have paraded Megawatt ratings like trophies because having the bigger number meant you had a better facility. This isn’t always the case.
In 2026, we need to discuss what happens when bragging about sheer size leaves you unable to deliver where it actually matters.
Facilities across the UK risk becoming what’s called “stranded assets”, where they sit on massive power contracts and have got the flashy megawatts, but they physically can’t keep modern hardware cool enough to run. Size is simply an illusion. Building a 50 MW facility tells you zilch about whether it can run modern hardware.
Historically, the commercial value of a data centre was tied to its utility grid agreement, so they would operate under the assumption that a 50 MW site would be twice as capable as a 25MW site; a dangerous game, for sure. Today, you can’t go by these assumptions anymore because they’re simply not true and thus lies the big megawatt myth: just because it’s huge, it doesn’t mean it operationally delivers.
The explosion of AI has increased performance and cooling demands. According to Data Centre World, in “The AFCOM State of the Data Centre Report: Insights and Trends for 2026,” AI accounts for 15% of workloads, but by 2030, it’s projected to rise to 40%. Rack densities have increased from 10kW-16kW up to 40kW, 80kW.
Hell even 100kW+ per cabinet, and when you pack that much compute into a small space, the power becomes less of an issue and the primary bottleneck becomes cooling solutions that efficiently and rapidly remove thermal energy away from the hardware and out of the room.
This creates stranded capacity, so an operator may have a 20MW power contract, but say if the facility just relies on raised-floor air cooling, then the facility will hit the thermal ceiling per se, before drawing even half of that power allowance. The other half of that expensive power allocation is just chilling, idle, stranded on paper, while enterprise tenants walk away to search for facilities with capacity.
The cooling crisis
Traditional facilities are hitting the wall because of how we cool buildings; air cooling relies on pushing cold air through floor grates, which works well if you have the space and when servers are spread out. However, air isn’t the best thermal conductor, so when servers are tightly packed together, air warms instantly and can’t pull the heat away fast enough, so the chips would fail or throttle.
Liquid is an excellent thermal conductor, transferring heat around 3,000 times more efficiently than air, so no wonder operators are investing heavily in liquid cooling solutions. Going from air cooling to liquid cooling also requires consideration of the physical architecture of the facility.
Operators have to consider structural floor loading because liquid racks weigh a lot more than traditional cabinets, so facilities built with lightweight raised floors now face structural limits.
There’s also the plumbing and hydraulics to factor in as electrical switchgear rooms are giving up floor space to Coolant Distribution Units (CDUs), secondary pipe loops and fluid pumps.
And finally, heat reuse integration, and if there’s the potential to reuse waste heat and supply it to local municipal district heating schemes, which is now a growing requirement in European urban data centres.
The UK reality is to upgrade or fall behind
Retrofitting an existing infrastructure and upgrading the cooling while maintaining tenant Service Level Agreements (SLAs) is like juggling whilst sat on a unicycle balancing on a tightrope. There is so much involved, from installing overhead piping, upgrading floor weight tolerances and introducing hybrid cooling solutions that serve standard air-cooled enterprise cabinets alongside liquid pods.
The purpose-built greenfield sites, on the flip side, are throwing away the legacy designs entirely and crafting blueprints specifically adapted to modern hardwire and cooling requirements. They are being built with no raised floors, closed-loop fluid architectures to eliminate excessive local water consumption and for thermal efficiency targets as operational benchmarks are shifting away from just PUE to holistic thermal capture metrics.
Here’s what facility leaders should be doing
Reframing procurement metrics is now top of mind. Leaders should stop pitching or evaluating solely on total megawatt capacity and instead, ask engineers for the maximum supported density per rack (kW/rack) and fluid flow capacity (L/min).
Conducting structural and hydraulic audits and identifying which rooms in your existing UK portfolio can support CDU retrofits and which ones can’t, so you can selectively choose which infrastructure is viable and plan for hybrid environments by building operational flexibility into upcoming fit-outs, so air cooling and liquid cooling can co-exist without integrating with the thermal dynamics of the other.
The era of bragging about megawatt capacity is over; size might get noticed on paper, but thermal intelligence and actual capacity being used will keep the lights on and the racks running efficiently.
