Energy Hungry Data Centres: The Power Debate Heats Up
The way we use electricity has evolved significantly, from simple light bulbs to electric cars, heat pumps and constant internet usage. This shift to a digital world, as the population grows, has led to a compounding increase in power demand.
Meeting this demand requires innovative strategies, including ramping up production and optimising existing power usage.
For the first time in 30 years, multiple nuclear projects are being considered in the US, and renewable energy sources are steadily growing.
However, a key strategy lies in also using the power we already have more efficiently.
Opportunities for improvement
Data centres – due to their high energy consumption rates, potential for heat recovery and AI-enabled capabilities – are well-positioned to impact energy usage positively.
According to a 2024 US Department of Energy study, domestic energy usage from data centres is projected to triple by 2028, accounting for 12% of all power.
Therefore, improving efficiency in data centres is crucial for reducing overall power consumption.
This can be achieved through optimised system design, components and control systems.
Regulators are pushing for more efficient Power Usage Effectiveness (PUE) rates, and heat recovery is a significant lever for any system with concurrent heating and cooling loads.
This can help attain higher Coefficient of Performances (COPs) than standard systems, making a substantial impact on direct power usage and lowering power requirements through heat recovery techniques.
Addressing wasted energy
Data centres and AI are major drivers of increased energy demand, but the heat they reject can be part of the solution.
Central Processing Unit (CPU) and Graphics Processing Unit (GPU) chips in data centres generate significant heat that must be removed to maintain system operations.
While various cooling systems exist, most reject heat into the atmosphere.
Existing data centres typically have minimal and inconsistent heating loads, making them unlikely candidates for much heat recovery on their own. Leveraging rejected heat requires collaboration with external applications or district heating plants.
Incentivising data centres and city planners to adopt a collaborative approach is essential for reusing waste energy effectively.
Strategic planning for heat recovery
Most data centres are not equipped for heat recovery. Hyperscale data centres, located outside metropolitan areas, are impractical for traditional heat recovery due to the high costs and heat loss associated with running miles of pipe.
The real opportunity lies in smaller, ‘edge’ data centres near potential loads, enabling collaboration with localities.
Stakeholders need to consider the number, scale and locations of data centres strategically. Communities are increasingly concerned about the water and power consumption of data centres.
Utilising heat recovery can reduce both water usage and power demands on the grid.
Harnessing heat effectively
Heat is effectively transferred through water systems in large cooling systems. Data centres can recycle low-grade waste heat using water-cooled chillers with a 4- or 6-pipe design to send water to a heat load.
For systems requiring hotter water temperatures, a high-temperature heat pump can boost the water temperature. This strategy pairs well with district heating applications, optimising heat recovery for entire communities.
Energy storage, combined with heat recovery, allows energy to be utilised on demand and stored for future use.
Managing peak power demand
Blackouts during heatwaves highlight the challenge of accessing power when needed.
Future power demand requires handling peak loads, which are amplified by complex AI calculations.
Stored energy can also help supplement power needs to meet peak loads. Thermal batteries, geothermal applications and other energy storage methods could be utilised for strategic power storage.
Europe leads in energy storage capacity, with 89 GW reported by the end of 2024. In the US, the Department of Energy (DOE) offers Technical Assistance Partnerships (TAPs) to help large energy users, including data centres, implement energy storage strategies.
The path forward
Reducing additional power needs starts with using existing power more efficiently.
Heat recovery and energy storage can significantly reduce overall and peak energy demand, providing more resilient power grids.
Designing integrated systems today ensures sustainability for future generations. AI isn’t just the reason for increased power – it is the key to pushing collaboration to harness necessary power.
