AI-Focused Data Centre Electricity Consumption Surged 50% in 2025. Here Is What That Means for Cable Management Specification.
- 10 hours ago
- 11 min read
AI-focused data centre electricity consumption surged 50% in 2025. Global data centre power use is set to double by 2030. The UK's share of national electricity consumed by data centres has already reached 5.9%. Behind every server rack, GPU cluster, and cooling system driving that surge is a cable management infrastructure that determines whether the facility performs reliably for the next 25 years — or starts accumulating maintenance problems within three.
Published by Reinforce Technology | May 2026
In April 2026, the International Energy Agency published its report Key Questions on Energy and AI — the most comprehensive analysis yet of what the global AI boom means for electricity systems. The headline numbers are striking. Electricity demand from data centres rose 17% in 2025, significantly outpacing global electricity demand growth of 3% (IEA, 2026). AI-focused data centres grew faster still, surging 50% in the same period. The capital expenditure of just five technology companies exceeded $400 billion in 2025 and is set to increase by a further 75% in 2026 — a figure that now exceeds global investment in oil and gas production combined (IEA, 2026).
The IEA's central projection sees global data centre electricity consumption doubling from 485 TWh in 2025 to 950 TWh by 2030, with AI-focused data centres tripling their power use over the same period (IEA, 2026). In the UK specifically, data centres already account for 5.9% of national electricity consumption — well above the global average of 2% and approaching the threshold at which the International Data Center Association warns that significant community and political resistance begins to materialise (Eastern Eye, 2026). UK government projections suggest that demand could quadruple by 2030 as AI adoption continues to accelerate.
These numbers describe an infrastructure construction programme of extraordinary scale and pace. Data centre developers are building AI factories — facilities specifically engineered for the power density, cooling intensity, and electrical complexity of GPU-based AI training and inference workloads. The capital expenditure per square metre of these facilities substantially exceeds that of conventional data centres. The electrical loading per rack is multiple times higher. The cooling infrastructure is correspondingly more complex. And the cable management systems routing power, data, and control cables throughout these facilities are under demands that conventional data centre design assumptions were never built around.
For every gigawatt of AI data centre capacity being built, there are kilometres of cable trays carrying the electrical infrastructure that makes it function. The specification of those cable trays — the material, the resin system, the structural performance, and the non-conductivity — determines whether that secondary infrastructure performs across the 25-year design life of the facility or becomes a source of corrosion-driven maintenance liability within years of commissioning. In a facility where unplanned downtime is measured in tens of thousands of pounds per minute, that is not a secondary procurement consideration.
Why AI Data Centres Create Uniquely Demanding Cable Management Conditions
Conventional data centres have always presented challenging secondary infrastructure conditions — elevated humidity from cooling systems, high electrical loads, and the absolute operational intolerance for unplanned downtime. AI data centres intensify each of these challenges significantly, and introduce new ones.
Power density is the defining characteristic of AI infrastructure. A conventional data centre rack might draw 5 to 10 kilowatts. An AI training rack — packed with GPUs — can draw 50 to 100 kilowatts or more, with next-generation configurations pushing significantly beyond that. Between 2020 and 2025, the power density of AI servers increased by eleven times, and a further fourfold increase is projected by 2027 (IEA, 2026). This density creates proportionally higher electrical loading on cable management infrastructure — more cables, higher voltage DC distribution, and greater thermal loads in cable tray zones.
Cooling systems are correspondingly more intensive. Liquid cooling — direct liquid cooling, immersion cooling, and rear-door heat exchangers — is increasingly deployed in AI facilities to manage the thermal output of high-density GPU racks. These systems introduce persistent elevated humidity, condensation cycling, and in some configurations, the risk of coolant contact with secondary infrastructure. In these zones, galvanised steel cable trays begin to corrode at exactly the fixing points and cut edges where they are most structurally important.
The electrical environment is more complex. AI data centres operate with both high-voltage AC distribution and high-voltage DC bus systems for GPU power delivery. DC arcing is more persistent and harder to interrupt than AC arcing. In this environment, conductive cable management infrastructure — steel trays that can become accidental current paths in the event of a cable insulation fault — introduces a risk category that non-conductive FRP cable trays eliminate entirely.
The rapid growth in AI data centre construction, driven by investment at a pace the IEA describes as unprecedented in energy history, is creating the largest single new market for advanced cable management specification in a generation. The facilities being built now will be operational into the 2050s. The cable tray specification made during construction determines their maintenance cost profile for that entire period.
The UK Data Centre Landscape in 2026
The UK's data centre market is growing faster than most European competitors. At 5.9% of national electricity consumption, the UK already has one of the highest data centre concentration levels in the world — above the 5% threshold at which the International Data Center Association warns political and community resistance accelerates (Eastern Eye, 2026). The queue for grid connections to the transmission network grew by 460% in the six months to June 2025, with multi-year waits now the norm for new projects seeking grid access (Energy Costs, 2026).
The government's response has been to accelerate rather than constrain. The UK Compute Roadmap identifies AI Growth Zones — nationally significant sites capable of serving at least 500 MW of demand by 2030, with at least one scaling to more than 1 GW. Data centres have been prioritised for grid connection alongside AI Growth Zones and EV charging infrastructure. The government's stated objective is to make the UK a leading location for AI data centre investment — and the infrastructure construction programme required to deliver that ambition is already under way at pace.
The investment flowing into UK AI data centre construction — as part of the global $400 billion-plus annual data centre capital expenditure programme — is creating a sustained pipeline of new facility construction that EPC contractors, M&E specialists, and infrastructure suppliers are delivering against timelines that leave little room for specification rework or material substitution. The decisions being made at design stage on each of these projects will determine their operational cost profile for the next quarter century.
Why FRP Cable Trays Are the Specification for AI Data Centres
1. Non-Conductivity in High-Density Electrical Environments
FRP cable trays are electrically non-conductive. In an AI data centre operating with high-voltage DC distribution to GPU racks, high-voltage AC distribution throughout the facility, and the elevated risk of cable insulation stress from high thermal loads, this non-conductivity is a direct and operational safety specification advantage. FRP cable trays cannot become accidental current paths. They require no earthing or bonding programme. There is no galvanic corrosion risk at the interface between the tray and other materials. And in the event of a cable insulation fault, the non-conductive tray provides no path for fault current propagation through the cable management structure (IntechOpen, 2022).
The earthing and bonding programme that steel cable management requires in a large AI data centre — with kilometres of tray runs across multiple high-voltage zones — is a meaningful element of the total installation cost and a recurring compliance verification burden over the operational life of the facility. FRP eliminates that programme element at installation and maintains that elimination across the facility's full 25-year design life.
2. Corrosion Immunity in Cooling-Intensive Environments
Liquid cooling systems in AI data centres — whether direct liquid cooling, rear-door heat exchangers, or the increasingly prevalent immersion cooling configurations — introduce persistent elevated humidity, condensation, and in some zones, the risk of coolant contact with secondary infrastructure. Galvanised steel cable trays in these environments begin to corrode at fixing points and cut edges within years of commissioning, as condensation cycling and elevated humidity attack the zinc coating at every location where it has been disrupted during installation.
FRP cable trays do not corrode. The corrosion immunity of FRP is not provided by a surface coating that can be breached — it is an intrinsic property of the composite matrix, consistent throughout the full cross-section at every cut edge and every fixing point. In the cooling zones of an AI data centre, FRP cable trays deliver identical structural performance across a 50-year design life without any corrosion-related maintenance intervention (Iris Publishers, 2020).
For an AI data centre operator whose facility represents hundreds of millions of pounds of capital investment and whose operational uptime is a direct revenue metric, the elimination of corrosion-driven maintenance access to live electrical zones is a directly quantifiable operational benefit.
3. Weight — Managing Structural Loading in Dense Overhead Installations
AI data centres are dense. Cable trays run in multilayer overhead configurations above hot aisle containment systems, through cable corridors between halls, and in high-density runs above GPU row infrastructure. The cumulative dead load of steel cable management at the density required for an AI facility is significant — and every kilogram saved in secondary infrastructure is either a kilogram that can carry additional cable load, or a kilogram that reduces the structural specification required for the ceiling and support systems.
FRP cable trays are approximately 75 to 80% lighter than equivalent steel sections (NACE International, 2016). In a large AI data centre fit-out, this weight saving compresses the installation programme — lighter sections handled by single operatives without mechanical assistance, faster fixing, no hot work permits — and reduces the dead load on the structural ceiling system. On projects where programme delivery speed has direct commercial value — and in AI data centre construction, every day of programme delivery delay represents a day of delayed revenue from a facility with very high committed operating costs — the installation speed advantage of FRP cable management is a quantifiable benefit from day one of the fit-out.
4. Thermal Non-Conductivity — Relevant in High-Load Cable Zones
FRP is thermally non-conductive. In cable tray zones carrying the high-density power cabling of AI GPU infrastructure, the thermal non-conductivity of the tray means it does not transfer heat laterally through the cable management structure in the way that steel does. This does not replace cable derating calculations or the thermal management design of the facility — but it is a consistent marginal advantage in the highest-load cable zones of an AI data centre, where every element of the thermal management design contributes to the reliability profile of the installation (IntechOpen, 2022).
The Sustainability Dimension: AI's Carbon Footprint and Infrastructure Materials
The energy and environmental footprint of AI infrastructure has become a high-profile issue alongside its growth. Data centres accounted for around 40% of all corporate power purchase agreements for renewables signed in 2025, and the pipeline of conditional offtake agreements between data centre operators and small modular reactor projects has grown from 25 GW at end-2024 to 45 GW in early 2026 (IEA, 2026). Technology companies are making large-scale commitments to zero-carbon electricity supply for their data centre operations.
The embodied carbon of the infrastructure itself — the materials used to build and fit out the facilities — is a less-discussed but increasingly relevant dimension of data centre sustainability. The UK Net Zero Carbon Buildings Standard, launched March 2026, requires embodied carbon to be documented and reported across new construction, creating a compliance framework that will increasingly apply to data centre construction as the standard extends beyond buildings into infrastructure assets.
FRP manufacturing emits 60 to 70% less CO₂ per tonne than traditional steel production (ScienceDirect, 2025). FRP cable trays weigh 75 to 80% less than steel equivalents, reducing transport emissions. They require no maintenance-driven replacement over a 50-year design life, eliminating the repeat embodied carbon of recoating and structural replacement that steel incurs in corrosive environments. For data centre operators and developers making sustainability commitments that extend to the full construction footprint of their facilities, FRP cable management is a directly documentable contribution to a lower embodied carbon outcome.
The Lifecycle Cost Case for AI Data Centres
AI data centres represent some of the highest-value built assets in the UK. A large hyperscale facility can represent £500 million to £1 billion or more of capital investment. The operational revenue model — selling compute capacity to AI developers, cloud services, and enterprise customers — depends entirely on uptime. Downtime in an AI data centre is not a production interruption. It is a direct revenue loss, a contractual SLA failure, and a reputational event in a market where alternative compute capacity is increasingly available.
In this context, the cost calculus for cable management specification is straightforward. FRP cable trays carry a higher upfront material cost than galvanised steel — typically 1.5x to 2x. Against a total facility capital cost of hundreds of millions of pounds, the cable tray material premium is a small fraction of the total project budget. The operational cost of accessing live electrical zones in an AI data centre for corrosion-driven cable tray maintenance or replacement — with the associated downtime risk, safety planning, and permit-to-work overhead — is disproportionately large relative to that premium.
The break-even point — where FRP's lower maintenance cost has fully offset its higher purchase price — typically falls within 8 to 12 years of installation (Younis, Ebead and Judd, 2018). For an AI data centre with a 25-year design life, the majority of the operational period is spent in net positive territory for FRP. And unlike a chemical plant or water treatment works where the maintenance events are planned and scheduled, a maintenance event in a live AI data centre carries the additional risk of operational disruption that cannot be fully planned around. The case for eliminating that maintenance category entirely — through the right cable tray specification at design stage — is stronger in an AI data centre than in almost any other infrastructure context.
Reinforce Technology FRP Cable Trays for Data Centres
Reinforce Technology supplies FRP cable tray systems for data centre applications across the UK and internationally, including AI data centres, hyperscale facilities, colocation campuses, and edge computing installations. Our range covers ladder trays, channel trays, perforated trays, and solid bottom trays in standard lengths and custom configurations for both overhead and below-floor cable management applications.
We work with data centre developers, M&E contractors, EPC contractors, and procurement teams across the UK's AI and hyperscale data centre construction pipeline. Contact us for technical data sheets, load and span tables, and project-specific specification guidance.
As with any infrastructure material, final confirmation of suitability for a specific data centre application remains the responsibility of the appointed project engineer or electrical designer. Reinforce Technology provides technical guidance and material recommendations based on the information supplied to us, but specification sign-off should always sit with the qualified professional responsible for the design. We are happy to provide full technical data sheets and application-specific support to assist with that process.
References
Brookings Institution (2026) Global Energy Demands Within the AI Regulatory Landscape. Available at: https://www.brookings.edu/articles/global-energy-demands-within-the-ai-regulatory-landscape/ [Accessed: May 2026].
Eastern Eye (2026) AI Explosion Pushes UK Datacentres to Consume Massive Share of National Electricity. Available at: https://www.easterneye.biz/uk-datacentres-electricity-ai-demand/ [Accessed: May 2026].
Energy Costs (2026) Will AI Data Centres Push Up UK Electricity Bills? Available at: https://www.energycosts.co.uk/articles/ai-data-centres-electricity-bills [Accessed: May 2026].
IEA (2026) Key Questions on Energy and AI. Paris: International Energy Agency. Available at: https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary [Accessed: May 2026].
IEA (2025) Energy and AI — Energy Demand from AI. Paris: International Energy Agency. Available at: https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai [Accessed: May 2026].
IntechOpen (2022) 'Fibre-Reinforced Polymer (FRP) in Civil Engineering', in IntechOpen Engineering Series. Available at: https://www.intechopen.com/chapters/84203 [Accessed: May 2026]. [GFRP confirmed electrical insulator; non-conductive properties in electrical environments].
Iris Publishers (2020) 'Global Impact of Corrosion: Occurrence, Cost and Mitigation', Green Journal of Earth Sciences, doi: 10.33552/GJES.2020.03.000618. Available at: https://irispublishers.com/gjes/fulltext/global-impact-of-corrosion-occurrence-cost-and-mitigation.ID.000618.php [Accessed: May 2026].
JLL (2025) Global Data Centre Outlook 2025. Chicago: Jones Lang LaSalle. Available at: https://www.jll.co.uk [Accessed: May 2026].
NACE International (2016) International Measures of Prevention, Application and Economics of Corrosion Technology (IMPACT). Houston, TX: NACE International. Available at: http://impact.nace.org/economic-impact.aspx [Accessed: May 2026].
ScienceDirect (2025) 'Sustainable composites for metal replacement: Environmental assessment and material selection of fiber-reinforced polymer across industries', ScienceDirect. doi: 10.1016/S2667-3789(25)00051-3 [Accessed: May 2026]. [FRP manufacturing emissions 60–70% lower per tonne vs steel; lifecycle carbon advantage documented across multiple independent studies].
Younis, A., Ebead, U. and Judd, S. (2018) 'Life cycle cost analysis of structural concrete using seawater, recycled concrete aggregate, and GFRP reinforcement', Construction and Building Materials, 175, pp. 135–144. doi: 10.1016/j.conbuildmat.2018.04.183. [~50% lifecycle cost saving for GFRP vs steel over 100-year study period].
Comments