As demand for data centres surges to keep pace with the increasing reliance on supercomputing enhancements and AI adoption, meeting sustainability objectives remains prominent on the agenda for businesses and the wider industry.
A hot topic in the market is the emergence of alternative cooling solutions to deal with the heat generated by the expanding scale of data centres, particularly those catering for high-density workloads driven by AI. Keeping data centres and its components cool is critical to optimizing efficiency, performance and longer-term reliability of hardware; not doing so can result in fundamental failures of the key systems and resilience. Balancing energy efficiency and sustainability with the development of high capacity and high-density workload data centres is also contributing to driving innovative cooling alternatives to traditional processes that consume large quantities of water.
Data centres embrace a number of different cooling technologies, dominated by traditional air cooling strategies, which circulate cold air from the natural environment/ air conditioning systems and expel generated heat. These are being challenged by the more heat intensive, higher-capacity modernised data centres. Whilst liquid cooling as a concept is not new to the market, there are increased discussions around exploring and adopting this method to more effectively and precisely transfer the higher heat generated by the equipment and servers contained in an AI facility. Increasing energy efficiency, driving down operation and maintenance costs, and reducing water consumption (particularly in geographical areas where water scarcity is a concern) are key factors in selection of the appropriate cooling method. Considerations as to how to make use of excess heat (eg to convert to a source of power) are also playing an interesting part in the conversations around energy efficiency/ sustainability of data centres.
There has been recent buzz and conflicting views about water consumption by data centres. Thames Water has warned of the pressure that new data centres will place on its infrastructure. Anglian Water has expressed similar concerns around supply due to regional water scarcity and potential flood risk. Voices from the data centre industry suggest that issues around water consumption and impact are oversimplified and do not take account of measures and technologies implemented by data centres. techUK published a report this month that found that 51% of surveyed data centre sites in the UK use waterless cooling systems and its findings challenge assumptions around heavy water-consumption, with facilities implementing efficiencies in the industry. Diverging opinions on data centre water use may be, in part, the result of a need for further data-driven policy making, reporting on water use and further investment to drive efficiencies in both data centre and water sectors, as highlighted by this report.
Sources of reliable power are vital to support the development of data centres, particularly AI facilities, and the electricity consumption of the industry has been the subject of similar discussions. The role of nuclear technology (in particular, small modular reactors) are increasingly being considered in the measures to provide secure power to data centre growth. Google has created strategic partnerships in the US to use SMRs to produce the energy that is required to feed its AI data centres, with other technology firms also considering nuclear energy to generate a secure and consistent electricity supply. With added advantages including low carbon emissions, cost stability, and enhanced land use efficiency (as advanced reactors can be designed with a smaller footprint in close proximity to the data centre facility), an increase in the use of SMRs and nuclear capacity more generally have been a feature of recent government announcements around energy security in the UK. It is not without its issues – high capital costs, relative infancy of the nuclear supply chain and acceptance of nuclear as an energy source in the UK do pose some potential obstacles. However, the urgency to build data centres (and associated infrastructure) and to create low carbon systems may well encourage the development of nuclear plants for data centres.
Accommodating increasing energy and AI demands are certainly contributing to the leap in development of new, modern data centres, including future proofing those assets to deal with technology advancements. So too, focus is turning to innovative ways to upgrade existing data centre facilities and retrofit energy efficiency/savings solutions to reduce costs and optimize sustainability and performance outputs.
