Technological strategies Several technological strategies can help enable sustainable AI: –Energy-efficient hardware (e.g. chips) and models reduce electricity consumption throughout the AI life cycle. –Innovative, insulated building materials reduce the need for heating, ventilation and cooling (HVAC) efforts. –Data centre infrastructure management software optimizes electricity use, improving system operation and maintenance. –Advanced cooling techniques can reduce consumption, compared to traditional methods. Featured technological use case TABLE 2 Virgin Media O2: AI-powered cooling optimization Situation/context Approach Results Virgin Media O2 partnered with EkkoSense to improve data centre efficiency.Virgin implemented EkkoSense’s AI- enabled approach to optimize thermal, power and capacity performance across 20 data centres.Benefits included cooling savings worth over £1 million per year, a 15% cooling electricity reduction and a 760 tonnes of CO2 saving. Operational strategies Several operational strategies can also support sustainable AI: –Incorporating target end use (model development versus training versus deployment) into site selection helps optimize efficiency based on workload. –Using scalable building designs that grow as demand increases mitigates oversizing. –Virtualization techniques reduce physical server requirements and consumption. –Temperature optimization and humidity management reduce overcooling and consumption. –Dynamic power management adjusts processing based on workload, reducing consumption. Featured operational use case TABLE 3 SAP: Aiming for “green” data centres Situation/context Approach Results Green data centres are key to SAP’s sustainability strategy.SAP data centres track resource use and minimize waste by using thermal cameras to optimize airflow and insulation, while also implementing cool/hot aisle containment to save energy.In 2023, SAP achieved carbon neutrality and is now on track to achieve net zero along its value chain by 2030.Source: Community consultation. Source: Community consultation. Artificial Intelligence’s Energy Paradox: Balancing Challenges and Opportunities 10

Artificial Intelligences Energy Paradox 2025