DATA CENTRES


The AI power crunch: how UK data centres are tackling unprecedented electrical demands


Tim Andrews, Janitza UK country manager


Roshan Rajeev, Vice President of engineering at Janitza USA


T


David Gilligan, VP critical power solutions & technology - global at Janitza Electronics


With the world’s biggest companies announcing major AI data centre expansions across the UK, the headlines are publicising job creation and technological leadership, yet behind the scenes engineers are asking a more pressing question: can Britain’s electrical infrastructure cope?


he country faces a near-20% surge in data centre capacity, with over 100 new projects in the pipeline alongside its existing 450+ sites. This explosive growth, fuelled by soaring AI computing demands, is spawning clusters around London and the Thames corridor, as well as emerging hubs in Manchester, Leeds, Wales, and Scotland. The answer reveals uncomfortable truths about the collision between AI ambitions and power grid reality. Across the UK, data centre operators are discovering that artificial intelligence workloads don’t just consume more electricity; they consume it differently. This creates power quality challenges that threaten equipment reliability, grid stability, and the viability of aggressive AI deployment timelines.


A new electrical landscape


Traditional data centres are predictable. Server racks hum along at steady loads, cooling systems maintain consistent draw, and power consumption follows recognisable daily patterns. But AI has shattered that predictability. “The electrical behaviour we’re seeing with AI workloads is fundamentally different.” explains Roshan Rajeev, vice president of engineering at Janitza USA, who previously managed data acquisition and analytics at Meta’s hyperscale facilities. “Model training creates sustained loads in the megawatt range; massive base loads that stress utility infrastructure. But inference operations are even more challenging from a power quality perspective.” Inference, the process of running AI models to generate outputs, creates what engineers call “burst activity”. A facility might spike from baseline to peak consumption in seconds as thousands of users simultaneously query large language models (LLMs). These high power, short- duration surges (or overvoltages), sometimes increasing by 100 kilowatts within 10 seconds, generate voltage sags, transients, and flicker that propagate through electrical systems. The hardware compounds the problem. Graphics processing units and tensor processing units powering AI computation draw massive amounts of current in highly nonlinear patterns. This creates harmonic distortion, electrical


Read the latest at: www.bsee.co.uk


“noise” that can damage transformers, interfere with sensitive equipment, and reduce system efficiency. In colocation environments hosting multiple tenants, aggregate harmonic distortion at the point of common coupling can exceed safe thresholds, affecting all customers.


Britain’s infrastructure reality


These technical challenges arrive as UK data centres face significant grid capacity constraints. According to industry analysts, grid connection queues have stretched to unprecedented lengths, with some projects facing waits of five years or more. The National Energy System Operator (NESO) for the UK has warned that data centre electricity demand could increase sixfold by 2030, requiring network reinforcement.


For AI facilities, the situation is acute. A single


large-scale AI data centre can require 100MW or more, roughly the output of a small gas turbine plant. Distribution network operators, already managing connections for renewable


energy projects, struggle to accommodate these enormous, unpredictable loads without risking grid stability.


“The power quality challenges we’re seeing in AI data centres represent a step change from traditional facilities.” says David Gilligan, VP critical power solutions & technology - global at Janitza Electronics. “Operators must monitor electrical parameters that previously weren’t critical: transients as brief as 18 microseconds, voltage harmonics up to the 127th order, and rapid load fluctuations that can destabilise distribution networks. A solution such as, Janitza’s UMG 801 power analyser resolves these challenges, making it a foundational tool for next-generation data centre management.” The regulatory environment adds complexity. UK planning frameworks, designed for conventional development, struggle to accommodate the speed of AI deployment. Energy efficiency requirements under Building Regulations Part L and ESOS compliance create additional hurdles for operators trying to balance performance with sustainability commitments.


BUILDING SERVICES & ENVIRONMENTAL ENGINEER JANUARY 2026 23


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36