Schneider Electric recently released a report on the future power consumption of artificial intelligence (AI) data centers, urging policymakers to carefully guide the electricity usage of AI data centers. This study was published following the International Energy Agency (IEA) Global Energy and AI Conference last month, titled "AI and Power: A System Dynamics Approach." The report explores various scenarios related to power consumption associated with AI, emphasizing the potential strain that the continuously growing AI infrastructure may place on the power grid.
Image Source Note: Image generated by AI, licensed from Midjourney
With the rapid rise of generative AI, significant investments are pouring into high-performance and energy-intensive infrastructures. The report indicates that existing data center infrastructure already consumes substantial energy, and as AI becomes more prevalent, electricity demand is expected to rise further, raising concerns about the potential strain on the grid and its environmental impact.
Schneider Electric's model presents four different future scenarios: Sustainable AI, Growth Constraints, Unlimited Abundance, and Energy Crisis. While all four scenarios predict an upward trend in electricity consumption from 2025 to 2030, the trajectory thereafter varies based on different assumptions.
In the Sustainable AI scenario, Schneider anticipates that power consumption in the AI sector will increase from 100 terawatt-hours (TWh) in 2025 to 785 TWh by 2035. In this scenario, AI infrastructure and demand form a complementary relationship that promotes efficient resource utilization. Conversely, in the Growth Constraints scenario, AI growth is limited by power and infrastructure bottlenecks, with electricity consumption expected to reach 510 TWh by 2030.
The Unlimited Abundance scenario warns that rapid and unrestricted development of AI systems could lead to unsustainable resource utilization, with electricity consumption projected to surge to 880 TWh by 2030 and reach 1,370 TWh by 2035. The Energy Crisis scenario illustrates the conflict between the rapidly growing AI sector and electricity demands from other economic sectors, potentially leading to economic recession and energy shortages.
In the report, Schneider Electric offers three recommendations to promote sustainable AI: first, optimize next-generation data center infrastructure by adopting the latest cooling technologies and high-efficiency AI hardware; second, enhance the efficiency of AI models through techniques like model pruning and quantization; and finally, establish a robust governance framework to develop certification standards for sustainable AI practices, ensuring that policies and technologies complement each other. The goal of the report is to guide government and industry leaders to balance AI growth with environmental and economic sustainability.
Key Points:
🌍 Schneider Electric warns that the power demand of AI data centers needs careful guidance to prevent strain on the grid.
📈 Four future scenarios predict different growth trends in electricity consumption between 2025 and 2035.
🔧 Three recommendations are proposed: optimize infrastructure, enhance AI model efficiency, and establish a governance framework to promote sustainable development.
