Technology

The rapid growth of generative AI (GenAI) is accelerating digital transformation but also increasing the ICT sector’s environmental footprint. Training, fine-tuning, and large-scale cloud deployment require massive computing resources, generating high energy use and embodied emissions. Current cloud metrics (tokens) fail to reflect actual hardware use, energy consumption, or lifecycle impacts, limiting transparency and comparability. The lack of standardised, verifiable methodologies hinders the integration of sustainability metrics into AI governance and design. Advantages and innovations Our work will deliver a harmonised, validated method linking AI usage metrics to real-world energy, hardware, and emissions data. Integrated into our fruggr platform, it will give AI developers actionable insights to design leaner, more resource-efficient models. To ensure accuracy and industry adoption, the project will partner with AI hardware and infrastructure providers to run large-scale, real-world impact measurements on their systems. These improvements will cut environmental impacts while reducing operational costs through lower energy and hardware demands. The methodology will also resolve inconsistencies between existing calculators, building trust and policy relevance. This will directly support the European Green Deal, AI Act implementation, and a competitive, low-carbon AI ecosystem. Technical Specification or Expertise Sought Field of expertise/experience: AI Hardware Providers • Proven track record in designing or manufacturing GPUs, AI accelerators, or custom chips for AI workloads. • Ability to share technical specifications (TDP, FLOPS, memory bandwidth) and facilitate on-site or remote performance/energy testing. HPC/AI Server Manufacturers • Expertise in building high-performance computing architectures optimized for AI. • Experience in integrating and benchmarking large-scale AI systems. Cloud Infrastructure Operators • Operation of large-scale, AI-capable cloud environnements (public, private, or hybrid). • Experience with workload monitoring, energy metering, and infrastructure optimization. Supercomputing Centres • Operation of Tier-0/Tier-1 HPC facilities. • Experience running large AI training and inference workloads, and collecting detailed performance and energy data. Stage of development - Under development IPR status - No IPR applied Sustainable Development Goals - Goal 13: Climate Action; Goal 9: Industry, Innovation and Infrastructure; Goal 8: Decent Work and Economic Growth;Partner Sought Type of partners sought: We are seeking for a hosting company with AI materials and services to gives us the possibility to test the impact of different AI models we will develop. - AI Hardware Providers – Manufacturers or designers of GPUs, AI accelerators, and specialized computing equipment to provide technical specifications and enable real-world measurement campaigns. - High-Performance Server Manufacturers – Producers of HPC and AI-optimized servers to support environmental impact testing on representative architectures. - Cloud Infrastructure Operators – Providers of large-scale, AI-capable cloud platforms, ideally based in Europe, to validate the methodology in production environments. - Supercomputing and HPC Centres – Facilities with large-scale computing resources to run benchmark tests and analyze workload efficiency.