Executive Summary

• The UK is ‘home to one of the world’s largest portfolios of legacy data centres’ and operators ‘must prove how fast they can innovate to stay ahead in the new AI landscape’
• Sources claim that UK data centre capacity which is AI-ready range from 8% to 20% today. By 2030, the government says the UK will require at least 6GW of AI-capable capacity
• Modular design will be key – and developers and funders are speaking directly with modular AI data centre solution providers right now – while nuclear energy could be a ‘strategic supply-side solution for digital infrastructure’

Writing her data centre trends for 2026 in December, Claire Keelan, managing director UK at Onnec, a provider of IT infrastructure design, installation and operational support, left readers in no doubt of the reality check about to hit them.

“With the UK home to one of the world’s largest portfolios of legacy data centres, operators must prove how fast they can innovate to stay ahead in the new AI landscape,” Keelan wrote. “In 2026, we’ll see a surge in retrofitted data centres as operators rush to upgrade legacy sites to meet soaring AI demand.”

This journey to AI is not going to be an easy one. The plan, as far as the UK’s Compute Roadmap published last year outlines, is for the UK to require at least 6GW of AI-capable data centre capacity by 2030. This represents a claimed threefold increase on capacity currently available.

“Should the capabilities and adoption of AI accelerate, demand could exceed this baseline significantly,” the government warns. But have no fear. By 2030, there will be a core group of ‘nationally significant sites’, each capable of serving at least 500MW of demand, with at least one AI Growth Zone scaling to more than 1GW. “Ideally, each site will be designed to expand into the 2030s, so that we can add capacity as AI demand continues to grow,” the roadmap adds.

But what about the existing infrastructure? A lot of data centre capacity now is cloud-built for an AI world. Paul Morrison, chief experience officer at applied AI provider Stelia.ai, wrote in November that while the UK sits on approximately 1.2GW of operational data centre capacity, only 8% of it is truly AI-ready. The rest is architected ‘for a completely different era’.

“Retrofitting that footprint for AI is like trying to run an industrial factory out of a Victorian townhouse,” Morrison wrote. Much like those who argue that Moore’s Law is nearing the end of its life, current data centre design will no longer take more heat or electrons. “The UK cannot afford to treat this as a conventional data centre expansion problem,” Morrison added.

Keelan feels similarly. “Power and cooling will be complex, but cabling and network capacity will be the real bottlenecks,” she wrote. “Poor quality or overcrowded cabling limits density, throttles performance, and makes future upgrades almost impossible.”

Morrison takes a cue from another government report; the Nuclear Regulatory Review, which calls for a ‘radical reset’ of nuclear regulation in the UK. Specifically, he notes that the report makes energy and digital policies dovetail, positioning nuclear as a ‘strategic supply-side solution for digital infrastructure’ – and a local enabler of AI ecosystems.

The next wave of data centres, Morrison posits, won’t look like data centres at all, but ‘integrated power-compute cells… self-contained modules that combine power, cooling, GPU density, storage and networking into a single engineered unit.’

That may be the far-out future. A look at the here and now, in terms of what data centre operators are thinking – and worrying about – comes in the BCS Data Centre Truths 2026 report. This analysis is a little more optimistic, but the flashing lights are the same; a ‘widening gap between confidence in AI-driven growth and readiness to support it’.

The report claims that 20% of data centres are AI-ready today, with expectations to rise to 70% by 2030. 85% of respondents, however, expect constraints on AI-capable floorspace. “Pressure around power, floorspace and rack density continues to shape how quickly AI demand can be converted into deployable capacity,” the report confirms.

Nick Marlow, head of project management at BCS, gives the high-level analysis. Government conversations have demonstrated a plan for adoption ‘with a desire for data sovereignty.’ Additionally, Marlow notes: “We are seeing developers and funders speaking directly with modular AI data centre solutions to meet pre-2030 deadlines.”

Modularity is a cornerstone of Keelan’s argument too. Modular designs ‘capable of scaling power, cooling and compute density will be key’ – and if you’re planning on building a facility in 2030, you’ve got to be prepared to be quantum-ready. “Smart operators will invest early in high-grade structured systems that support modular expansion and long-term flexibility,” notes Keelan. “Retrofit-ready will become the new benchmark for responsible, future-proof design.”

Whether it’s about retrofitting, or ripping it all up and starting afresh, the industry is certainly living in interesting times right now.