Robert van der Kolk, President of EMEA and APAC at nVent, explains why the next phase of AI infrastructure will depend on rethinking power delivery, cooling and the way the two work together.
As AI usage expands, the hardware that supports it is becoming as critical to the AI revolution as the large language models themselves. While there are different kinds of data centres, from giant hyperscale projects to edge installations, all data centre operators are focused on the same things: deploying AI quickly, delivering and using power efficiently, and future-proofing IT infrastructure so it is prepared to support the needs of AI chips, graphics processing units (GPUs) and tensor processing units.
Power and cooling requirements
The scale of technological change happening in data centres is immense. For context, estimates put average pre-AI, non-high-performance computing data centre rack power at about 8 kW per rack. Today, the industry is pushing towards one- to three-megawatt racks, representing a substantial increase in power demand.
As AI applications are deployed across most industries, supporting AI technology requires a significant shift in infrastructure thinking.
Cooling reference architectures
Given these power demands, liquid cooling is becoming increasingly important for AI data centres, as air-based cooling technology alone is often insufficient to manage GPUs effectively. To implement suitable solutions successfully, the industry needs a flexible and collaborative framework for developing AI cooling infrastructure standards. Moving towards common infrastructure standards and a more interoperable model within the data centre industry could help companies accelerate development while continuing to innovate in differentiated ways.
Modularity and standard interfaces allow data centres to deploy technologies faster, providing a foundation upon which infrastructure providers can innovate for efficiency and performance. Reference architectures can still allow for unique and differentiated designs in coolant distribution units, rear-door coolers, heat rejection units and manifolds, and technology cooling systems, while also providing compatibility and interface standardisation that supports the mixing and matching of products from multiple suppliers.
The liquid cooling architecture used in today’s high-density racks is typically a closed-loop system that uses a treated fluid that is continuously recirculated. By leveraging conduction over convection, direct-to-chip cooling can prevent evaporative loss and improve efficiency compared with air cooling. This architecture not only supports thermal management but can also provide high-grade waste heat, creating opportunities for heat reuse applications across the data centre.
High-voltage DC power
For megawatt-scale rack power delivery, the industry is moving towards 800-volt direct current (VDC) power distribution and, potentially, 1500 VDC in the longer term.
This shift in power delivery architecture can offer a number of benefits, including reducing copper usage and minimising resistive losses. Bringing DC power all the way to racks also reduces the number of AC/DC conversions in a data centre. This can make installation easier and lower costs for data centre operators, while also reducing the power losses that occur whenever power is converted from one form to another.
Moving to DC power can also simplify complex data centre construction projects because electricity does not need to be converted and reconverted several times on its way to IT racks. This shift may also enhance the scalability of data centres because power distribution infrastructure does not need to be redesigned when IT infrastructure is added, only expanded to include additional racks.
However, data centre operators need to understand that products such as AC-to-DC converters, busbars and busways, and rack power delivery, protection and monitoring solutions will need to be redesigned from how they look today to fit into a DC power architecture.
Power and cooling convergence
Cooling and power infrastructure in data centres need to work in tandem. To achieve maximum output efficiency, GPUs need the right amount of power, and the resulting heat energy needs to be appropriately dissipated by cooling infrastructure to keep these chips within their thermal operating parameters.
Optimising cooling and power together is an area where data centre operators can improve both efficiency and performance. Given the increase in rack power and the resulting transients, a connective framework, advanced control algorithms and a software management layer between the IT equipment and the power and cooling infrastructure are likely to become increasingly important for managing infrastructure efficiently and safely.
AI’s next phase will be shaped in large part by the physical layer. Megawatt-class racks are likely to require liquid cooling built on interoperable reference architectures, alongside high-voltage DC distribution to help reduce losses, materials use and complexity. The opportunity now is to bring together power, cooling and software controls in a way that supports efficient, safe and predictable scale.
