From Paradox to Progress A Net Positive AI Energy Framework 2025

Page 30 of 38 · WEF_From_Paradox_to_Progress_A_Net_Positive_AI_Energy_Framework_2025.pdf

Embedding such measures into reporting and oversight systems enables comparability, accountability and continuous improvement in resource use. In turn, this creates a foundation for transparent, data-led progress towards more efficient AI. Key levers include: –Global energy efficiency metrics: Joules per inference, life cycle carbon intensity –Public disclosure frameworks: AI energy use ESG reporting standards –Benchmarking platforms: Cross-industry tools for comparing AI workloads –Third-party verification: Independent audits of AI infrastructure and models –Open data repositories: Shared datasets for energy impact research Use case insights and takeaways: A notable 81% of use cases incorporate this enabler. Many rely on dashboards or carbon- tracking systems. However, few extend to standardized reporting or external validation. Establishing global standards and open platforms will be essential to scale impact and build trust.Strategic recommendations: –Establish AI energy efficiency and carbon intensity global standards –Require benchmarking compliance for public procurement –Support open data platforms and third-party auditsTransparent measurement and accountability key levers FIGURE 12 Global energy efficiency metricsPublic disclosure frameworksBenchmarking platformsThird-party verificationOpen data repositoriesTransparent measurement and accountability Emerging transparent measurement and accountability use case examples TABLE 7 Source: AI Energy Impact public use case database.Global energy efficiency metricsNorth American information technology (IT) firm (automated server decommissioning): Monitoring identified underused servers with deactivation saving 10,475MWh and 3,506 tonnes of CO2, demonstrating life cycle energy accounting and efficiency benchmarking Public disclosure frameworksNorth American and European innovators (transparency and efficiency): Published model emissions and resource use data; created an AI energy score leaderboard, setting benchmarks for public reporting and accountability Bench- marking platformsARCEP , France (ICT sustainability benchmark): Developed a national platform to measure and publish energy and emissions data across digital operators, creating transparent benchmarks for environmental accountability Third-party verificationCartoBio, France: Improved product efficiency through feature retirement and interface redesign, reducing resource use and exemplifying digital sobriety in sustainable software design Open data repositoriesGlobal sustainability consortium: Curates public case studies enabling transparent cross-sector comparison and shared sustainability baselines From Paradox to Progress: A Net-Positive AI Energy Framework 30
Ask AI what this page says about a topic: