The rapid global adoption of AI is driving computational demand beyond the limits of existing digital infrastructure. As AI workloads grow, data centres are consuming increasing amounts of electricity—much of it generated from non-renewable sources—resulting in rising carbon emissions. Cooling these facilities also requires vast quantities of water, with global data-centre water consumption reaching an estimated 175 billion litres in 2023 and projected to more than triple by 2030.
To address these challenges, we are developing a new generation of high-bandwidth optical neuromorphic AI compute chips that deliver substantially higher computational efficiency while reducing energy and cooling requirements. Experimental results to date demonstrate up to 7× greater energy efficiency, 10× higher processing throughput, and improved thermal stability compared with conventional GPU- and TPU-based AI accelerators.
By significantly reducing power consumption and heat generation, our technology has the potential to lower both CO₂ emissions and water usage associated with large-scale AI infrastructure. With the global AI market projected to exceed £300 billion by 2027, this translational research offers a pathway to commercialising next-generation AI hardware that combines high performance with environmental sustainability, supporting the growing demand for greener AI computing.
