IN Brief:
- Fluke reports that 79% of UK data centre respondents experienced electrical issues contributing to operational problems or serious near-misses during the previous year.
- Only 17% consider their electrical infrastructure fully prepared for higher-density AI workloads, while 32% report power quality problems affecting deployment.
- Electrical skills shortages have delayed infrastructure upgrades for 38% of respondents, alongside reported limitations in system monitoring and grid connection capacity.
Electrical infrastructure problems are affecting UK data centre operations as facilities prepare to accommodate higher-density artificial intelligence workloads, according to research commissioned by Fluke Corporation.
The company’s study found that 79% of UK respondents had experienced electrical or power-system issues contributing to disruption, delays, downtime, or serious near-misses during the preceding year. Only 17% considered their electrical infrastructure fully prepared for higher-density AI workloads, compared with 33% of respondents globally.
Released on 17 November, the findings form part of an international survey conducted by Censuswide on behalf of Fluke. The research involved 1,300 data centre operators and industry professionals across the UK, United States, Germany, France, India, and Saudi Arabia.
The UK findings identify constraints within data centres alongside difficulties obtaining additional electricity from the network. Seventy per cent of respondents said access to sufficient power or grid capacity was moderately or severely restricting expansion, while the average reported wait for additional power capacity associated with AI and GPU developments was 14 months.
That waiting period was almost two months longer than the global average. However, securing additional capacity at the grid connection does not establish that the electrical infrastructure inside a facility can distribute the increased load to its computing equipment.
Incoming electricity must pass through a sequence of distribution and protection systems before reaching individual server racks. Depending on the facility’s configuration, these include transformers, high-voltage or low-voltage switchgear, uninterruptible power supplies, distribution boards, busways, protective devices, and local power distribution units.
Each stage has an electrical rating and operating limits that determine the current it can carry and the conditions under which it can remain in service. A connection capable of delivering additional megawatts does not remove restrictions imposed by downstream transformers, conductors, switchgear, or protective equipment.
AI computing systems can concentrate substantial electrical loads within comparatively small areas. Increasing rack power density may require additional distribution capacity, revised protection arrangements, different cooling provisions, and changes to the equipment supplying critical loads.
These modifications must be assessed against the existing installation. Cable and busbar ratings, transformer loading, short-circuit conditions, and equipment temperature limits can influence whether the distribution system can accommodate the proposed demand safely.
The research indicates that existing electrical performance is already affecting deployment decisions. Almost one-third of UK respondents, 32%, said power quality issues had reduced their ability to deploy or operate high-density AI or GPU workloads.
Power quality encompasses characteristics of the electrical supply that can influence connected equipment, including voltage disturbances, waveform distortion, and variations from specified operating conditions. The consequences depend on the disturbance, its duration, and the performance requirements of the affected equipment.
Electronic power supplies and other conversion equipment are designed to operate within defined electrical limits. Persistent or severe disturbances can affect their performance, while abnormal conditions elsewhere in the distribution system may require protective equipment to operate.
The survey does not identify one specific electrical disturbance as the principal cause of the reported problems. Its 32% figure concerns respondents who reported an effect on AI deployment or operation, rather than an independently measured frequency of voltage disturbances or equipment failures.
Detecting electrical problems before they interrupt operations requires suitable measurements at the relevant parts of the distribution system. Only 41% of UK respondents reported real-time visibility across their critical electrical systems, while 26% said problems were identified only after performance deterioration, equipment failure, or downtime.
Monitoring arrangements may include measurements of current, voltage, power consumption, equipment temperature, and operating status. Where instruments capture disturbances with appropriate resolution, their records can support investigations into transient events, changing load patterns, and abnormal equipment behaviour.
Measurements made at the incoming supply do not necessarily reveal every problem occurring downstream. A fault developing within a particular distribution assembly or electrical connection may require more localised monitoring or inspection before its cause can be established.
The usefulness of monitoring also depends on how recorded information is interpreted. Alarms and measurement trends can identify conditions requiring investigation, but they must be assessed against the design of the installation and the operating limits of the affected equipment.
Maintenance and electrical testing remain necessary even where extensive monitoring is installed. Circuit breakers, protective relays, cable terminations, switchgear, and other equipment require inspection and verification appropriate to their condition, duty, and maintenance requirements.
Data centres operate with limited tolerance for interruptions to critical computing loads. Maintenance work must therefore be coordinated with available supply redundancy, isolation arrangements, and the operating requirements of equipment that cannot readily be taken out of service.
Those activities require suitably qualified personnel. Fluke found that 51% of UK respondents considered recruitment of qualified electrical technicians more difficult, while 38% reported that skills shortages had delayed infrastructure upgrades or the introduction of additional capacity.
Electrical expansion projects can require specialist engineering across several disciplines, including high-voltage and low-voltage distribution, protection, testing, power quality, critical power systems, and commissioning. Equipment delivery alone does not resolve a shortage of personnel able to install and validate the completed systems.
Commissioning work must establish that new electrical equipment operates as intended and that it can be integrated with the existing distribution arrangement. This may involve insulation testing, protection verification, functional testing, and checks of supply transfer arrangements, depending on the installation.
Where existing facilities are being expanded, engineers must also determine whether the new equipment changes fault levels, protection coordination, or the loading of shared electrical assets. These requirements can affect both the design and the sequence of works needed to introduce additional computing capacity.
Fluke president Parker Burke identified the management of electricity already available inside data centres as an area requiring greater attention alongside grid access. The company advocates improved system visibility and proactive maintenance as operators prepare for increasing power demand.
The reported results are survey responses rather than a verified register of electrical failures. The 79% figure combines disruption, delays, downtime, and serious near-misses, and must not be interpreted as meaning that 79% of UK data centres suffered an electricity outage.
The published methodology confirms an international sample of 1,300 respondents across six countries, but the supplied UK findings do not specify the number of UK participants. The UK percentages therefore cannot be assigned an independently calculated national margin of error from the information available.
The comparison between the UK and international results nevertheless provides a consistent picture of reported infrastructure pressures. The UK recorded lower confidence in readiness for higher-density AI workloads, while respondents also described limitations in power quality, technical workforce availability, and monitoring.
Additional computing capacity will require operators to assess the condition and ratings of the electrical systems between the network connection and the equipment they intend to install. The findings indicate that internal distribution capability, maintenance resources, and grid access are already affecting the pace at which new AI infrastructure can be deployed.


