Energy efficiency and sustainability of AI data processing in Data Centres (IA)

Apie kvietimą

Expected Outcome: Pilots for new technologies including demonstration sites which contribute to energy-efficient and sustainable AI data processing in data centres, reinforcing EU strategic autonomy and climate goals. Project results are expected to contribute to one or more of the following outcomes: Demonstrated innovations that substantially improve heat removal from high-power AI chips (e.g. direct on-chip cooling, advanced thermal interface materials, multi-scale thermal management), enabling higher performance without thermal throttling. This should lead to lower cooling energy needs and higher reliability for dense AI workloads. Prototypes of novel backup power systems (such as graphene-enhanced batteries) that operate with minimal cooling requirements, improving data centre resilience and enabling better use of renewable power. New methods and frameworks that optimise the entire data centre for energy-efficient AI processing. This includes intelligent workload scheduling and AI model optimisation techniques to reduce energy use (e.g. carbon-aware job scheduling and power capping to cut energy demand and peak temperatures), as well as designs for integrating on-site/off-site renewables and waste-heat reuse. Outcomes should demonstrate potential for improved power usage effectiveness and utilisation of waste heat in external applications, aligned with European targets for carbon-neutral heating/cooling. Scope: Actions should address one or more of the following development areas: Direct on-chip cooling and thermal management, including novel and innovative cooling techniques applied at chip and module level (direct liquid cooling, heat spreaders, thermal interface materials, and advanced packaging) and multi-scale thermal management techniques. Energy-efficient power backup and storage systems: Innovations in early-stage energy storage concepts (graphene-enhanced batteries, supercapacitors, and other emerging battery chemistries) and approaches for net-zero backup. Sustainable data centre architectures and AI workload optimization: addressing AI-driven workload scheduling, adaptive power management, dynamic resource allocation and integration of data centre heat capture and reuse. Materials research for energy efficiency: Projects to make use of existing research in new materials and components supporting energy efficiency and thermal management, and to employ these for data centres benefit. Optimisation of data centre operation and functioning: explore AI solutions to optimize the Data centre functioning, computing architecture, and virtualization, minimizing its carbon and environmental footprint. Integration of data centres into energy systems and the wider region: including solutions that integrate Data centres into energy system planning and operation. In addition, actions should establish open pilot demonstration sites to test and integrate research results, and to demonstrate, benchmark, and disseminate findings among relevant…