High-performance computing (HPC) centres and data centre operators face a growing challenge: energy budgets are constrained, but the energy consumption of individual workloads is often only known after execution. As AI and compute-intensive workloads continue to grow, operators must improve energy efficiency, reduce operational costs, and meet increasingly demanding sustainability reporting requirements.
In Ireland, data centres already account for a significant proportion of national electricity demand, while across Europe regulatory frameworks such as the Energy Efficiency Directive (EED) Article 12 and Corporate Sustainability Reporting Directive (CSRD) are increasing pressure for transparent energy measurement and reporting. However, existing tools cannot accurately predict workload energy consumption before execution or determine the most energy-efficient infrastructure location for a given workload.
This project will develop an AI-driven pre-execution energy prediction and workload optimisation platform for heterogeneous data centres. Before deployment, the system analyses workload characteristics alongside real-time infrastructure telemetry across compute, memory, storage, network, and cooling systems to predict energy consumption across possible configurations and recommend the optimal placement balancing performance and energy efficiency.
The platform is underpinned by three key innovations: a runtime voltage prediction model that estimates processor behaviour from system telemetry; a physics-informed energy prediction engine covering CPUs, GPUs, AI accelerators, memory, storage, and network components without requiring additional power meters; and an optimisation engine that identifies the most efficient workload placement before execution.
The target customers are HPC centres and colocation providers across Ireland and Europe. Following pilot validation, the technology will be commercialised through enterprise software licensing, enabling operators to reduce energy costs, improve infrastructure utilisation, and meet emerging sustainability requirements.
