Demand for Artificial Intelligence (AI) has rocketed in the last few years. What only a couple of years ago was the preserve of academia and a few big tech firms, is now firmly entrenched in everyday lives.
To meet the rampant demand for data services and accommodate the increasing sophistication of AI, the world is embarking on a data centre boom. The number of data centres in the UK is predicted to grow by a fifth in the next five years. That pales in comparison to the nearly 3,000 data centres under construction or planned in the US, taking the total to 7,000. Worldwide, data centre investment hit $61 billion in 2025.
The UK’s government has thrown its weight behind data centres – a key component of its proposed AI Growth Zones, through which it “aims to establish AI innovation hubs across the UK”. However, it also has a mission to deliver Clean Power by 2030 and a legal obligation to reach net zero by 2050.
While the prolific increase in the number of data centres being planned brings their environmental impact into the spotlight, they could also be critical in supporting progress towards net zero carbon emissions and improvements in environmental sustainability.
Possible applications for AI range from optimising our energy consumption by managing demand on the grid, reducing waste, and monitoring the health of environmental ecosystems.
AI can also be a powerful enabler for sustainable innovation through for example the use of machine learning and digital twins to accelerate the identification of solutions. Additionally, AI could also help to quantify the impact of climate change and adaptation strategies.
In Great Britain, electricity demand is estimated at 7.6 TWh from connected data centres, according to the National Energy System Operator. This equates to all of the electricity generated by solar power over five months in the UK last year, representing a significant amount of the energy produced from renewable sources.
In Ireland, data centres are estimated to account for 21% of total electricity consumption.
The environmental impact goes beyond electricity consumption and the associated emissions too.
Data centres can consume significant volumes of water for cooling servers. This water tends to be withdrawn and consumed from nearby sources, exposing local communities to the risk of water scarcity. The Environment Agency’s own analysis projects that England faces a 5 billion litre per day shortfall in water supply by 2050.
But the environmental impacts could be much greater due to inconsistent accounting methods – with reports that alleged emissions could be as much as 662% higher than officially stated.
With such drastic environmental consequences, we urgently need to transform the energy efficiency and cooling needs of data centres to make AI more sustainable as it develops and its adoption grows across society.
Government policy and regulation have a very significant role to play here, as highlighted by the National Engineering Policy Centre’s calls for mandatory environmental reporting and requirements for data centres – but some wins could be achievable through research and innovation too.
The Royal Academy of Engineering has recently awarded £6 million to two innovators who aim to slash the environmental impact of data centres. The funding is part of a wider £39 million announcement of 13 new Green Future Fellowships, with a mission to scale a variety of solutions to the climate crisis into commercially viable ventures.
Dr Rostislav Mikhaylovskiy at Lancaster University is developing a new type of memory that uses extremely short bursts of terahertz radiation – light pulses a thousand times faster than today’s technology. These flip the direction of small magnets that store bits of data. Because these pulses match the magnets energy, they can switch them without creating heat. This could lead to much faster, cooler and more energy-efficient data storage in the future.
Meanwhile Dr Kilian Stenning, founder of Rayd Technologies, is scaling reconfigurable, non-linear photonic computing for energy efficient AI. He is developing brain-inspired “neuromorphic” computing that uses light (photons) to process data and images extremely efficiently, with potential for more than 10,000 times less energy than current GPU microchips.
The system can learn from small datasets where traditional AI falls short and already works for tasks like image classification and medical imaging.
By miniaturising the hardware and expanding its capabilities, this technology could drastically improve AI’s speed, data efficiency, and operational electricity consumption. This could help to make AI faster and more environmentally friendly and enable a new class of AI.
Cutting the energy demand of AI and data is critical to removing carbon emissions from the industry. Tackling heat generation can cut both the water and energy currently used for cooling.
Scaling solutions like those being developed by Dr Mikhaylovskiy and Dr Stenning into commercially viable technologies will enable their deployment in data centres around the world.
By adopting these innovations, we can not only lessen some of the most severe environmental impacts caused by data centres but also enjoy the advantages of expanding AI use in a more sustainable manner.
