Executive Summary
- Next-generation data centres will only succeed if they can scale aggressively while preserving the stability of the electrical ecosystems they depend on.
- For designers and builders, the challenge is to blueprint new, AI-ready infrastructure.
- All without disrupting the delicate balance of the existing digital matrix that keeps today’s workloads available, efficient and secure.
The new data centre reality
Modern data centres are no longer passive consumers of electricity; they actively shape grid behaviour through highly dynamic, software-orchestrated loads. AI and high-performance computing (HPC) clusters can ramp several megawatts of power in seconds, remain within IT equipment tolerance curves, and still stress the wider electrical system with voltage sags, harmonics and rapid load swings.
This means that traditional design assumptions, which focused primarily on capacity and redundancy, are no longer sufficient for the next generation of facilities.
At the same time, colocation and hyperscale operators face mounting pressure to improve energy efficiency, document power quality, and report against standards such as EN 50160, EN 61000-2-4, DIN EN 50600-2-2, ISO/IEC 22237-3 and ISO 50001.
For AI deployments in particular, the case for deeper visibility is becoming more urgent, because higher rack densities, rapid switching behaviour and more localised cooling architectures increase the need for measurement at finer levels of granularity.
Designing for power quality and resilience
In AI-centric environments, non-linear loads from GPUs, UPS systems and switch-mode power supplies create complex power quality profiles that fluctuate in milliseconds. These profiles drive phenomena such as harmonic distortion, voltage flicker and transient events, which can destabilise generators, overload UPS rectifiers and cause unexpected tripping of downstream equipment if they are not carefully monitored and managed. From a design perspective, power quality must be treated as a core architectural constraint rather than an afterthought addressed only during commissioning.
A proven approach is to implement a multi-level power quality architecture that spans from the utility connection down to the rack and final circuits. Janitza’s wider data centre guidance aligns this with three standard-defined granularity levels, allowing operators to measure primary and secondary supplies, intermediate distribution points and protected socket outlets with the degree of precision each layer requires.
Class A power quality analysers at the main distribution level provide holistic visibility of voltage quality, frequency behaviour and residual currents, while modular, MID-certified devices at secondary distribution and rack level offer granular insight into specific feeders, busbar systems and IT loads.
In practice, this often means deploying Class A power quality analysers at the main incomers and UPS outputs, with other analysers on critical feeders and low-voltage main distribution boards to capture harmonics, flicker and transient events in line with EN 61000‑2‑4 and EN 50160.
For dual-path architectures and catcher systems, a modular, all-in-one analyser can monitor two supply paths simultaneously, detect fast switching events down to the microsecond range and evaluate compliance with compatibility levels automatically.
Residual current monitoring as a design principle
Residual current monitoring (RCM) is emerging as a critical discipline in high-density, AI-ready facilities. Leakage currents from thousands of switch-mode power supplies, long cable runs and complex UPS topologies can accumulate over time, masking real insulation faults, increasing fire risk and triggering nuisance trips in protection devices. In conventional designs, these effects were often manageable; in a campus-scale AI facility with megawatt-scale clusters, they can become a systemic risk if left unmeasured.
Integrating continuous RCM at multiple granularity levels, including main distribution, sub-distribution and rack circuits, enables operators to distinguish between background leakage and genuine fault currents.
This is also significant from an operational standpoint, because standards for data centres require the integrity of TN-S systems to be safeguarded and because continuous monitoring can reduce the need for disruptive insulation testing during periodic inspections.
By embedding RCM into the design blueprint, data centre owners can future-proof their protection strategies against steadily rising leakage levels as capacity expands.
Energy transparency and sustainable performance
Sustainability expectations are reshaping how data centres are planned, built and operated. Operators must not only reduce energy consumption but also demonstrate continuous improvement through key performance indicators such as Power Usage Effectiveness (PUE), Energy Usage Effectiveness (EUE), and, where water cooling is involved, Water Usage Effectiveness (WUE). Achieving this requires far more than simple meter readings; it demands holistic energy transparency across the entire electrical infrastructure.
Granular metering at three levels, namely primary and secondary supply, distribution and IT final circuits, provides the data needed to calculate PUE, identify losses and benchmark efficiency over time.
For AI and HPC environments with very high rack power and local rack cooling, level 3 measurement becomes especially valuable because it allows operators to assess performance much closer to the actual load and cooling interaction. When combined with a dedicated power grid monitoring platform, these measurements can be turned into dashboards, automated reports and alarms that support both daily operations and long-term optimisation.
Blueprinting the digital matrix
The next generation of data centres will be defined as much by their architectural intelligence as by their physical footprint. Architecturally, this means designing measurement and control layers alongside power delivery and cooling from the earliest concept stages rather than retrofitting them later.
It also means embracing modular, scalable designs, such as segmented low-voltage main distribution boards, busbar trunking systems and intelligent PDUs, that can grow with demand while maintaining transparency and control.
In practice, a balanced blueprint might include Class A analysers at the main incoming supply, modular network analysers at secondary boards, and multi-channel operating and residual current monitoring at rack level, all unified within a single software environment.
That same architecture should also allow integration with higher-level SCADA, BMS or DCIM environments through protocols such as SNMP, Modbus, OPC UA and REST API, so that electrical insight becomes part of a wider operational decision-making layer rather than a standalone dataset.
Designers can then focus on integrating new technologies, such as higher-density racks, liquid cooling or on-site generation, without losing sight of power quality, safety and efficiency.
Emerging technologies and services
As electrical infrastructures become more complex, the combination of advanced hardware, intelligent software and expert services becomes a key differentiator.
Software platforms capable of integrating power quality events, residual current data, energy KPIs and environmental signals into a unified view give operations teams the situational awareness they need to manage highly dynamic loads.
Open connectivity via standards such as OPC UA and REST APIs further allows this data to be shared with SCADA systems, building management platforms and analytics tools for deeper insights.
Editor’s note: This article is taken from the Fall issue of DCi’s Frontiers magazine, and you can read the full, digital publication here.