Source: Stakeholder consultations.Challenges faced in sustainable AI applications TABLE 5 Challenge Challenge details Energy infrastructure and availability –Power reliability: Delayed utility upgrades to meet data centre demand can disrupt operations when required demand isn’t delivered. –Cooling needs: Increasing temperatures due to climate change led to heightened cooling demands, straining power during peak summer months. Data and computational –Data mobility and optimization: While data processing can move across locations, optimizing data workloads based on energy availability remains complex. –Data quality: Poor data quality can reduce AI’s energy efficiency, e.g. necessitating frequent retraining due to inefficient model performance. Regulatory and policy –Lack of standards: A lack of uniform global standards, taxonomies and definitions for AI energy use and digital systems creates challenges in scaling and assessing impact. –Regulatory complexities: Regional differences in regulations complicate compliance, especially as AI-focused regulations emerge. –Industry-specific regulations: Some regulations, (e.g. for building efficiency and renewable adoption) vary widely across industries and regions. Industry collaboration and partnerships –Complex value chains: Difficulty in mapping supply chains and gathering supplier data impacts collaboration. –Dependence on key players: Concentration of R&D within a few companies limits innovation accessibility. –Lack of local capacity: There is inadequate local infrastructure and an insufficient volume of skilled partners for implementing global AI systems. Additionally, there is a shortage of talent skilled in both AI and energy, limiting innovation and implementation speed. –Risk aversion: Telecommunication and energy sectors are hesitant to adopt disruptive technologies, focusing instead on incremental changes. Mindset, awareness, and cultural shifts –Awareness barrier: There is low awareness about the financial benefits of sustainable AI. –Mindset shift: Across supply chains, there is hesitation to adopt energy-efficient practices, often due to fears around disrupting established profit margins. –Geographic variations in attitudes: Different regions prioritize AI and sustainability goals based on their socioeconomic and environmental context Operational and technical challenges –Data privacy and security: As AI models require vast data, privacy and security concerns become a significant barrier, especially in regions with strict regulations. –Energy security concerns: Dependency on external energy sources and concerns about energy reliability due to geopolitical issues can create hurdles in AI applications. –Real-time forecasting and optimization: There is a need for accurate forecasting tools to optimize resource allocation and identify energy hotspot Artificial Intelligence’s Energy Paradox: Balancing Challenges and Opportunities 20

Artificial Intelligences Energy Paradox 2025