算力狂奔的尽头是“能耗悬崖”?他们正在寻找答案 | 两说
第一财经·2026-01-08 06:33

Core Viewpoint - The rapid evolution of AI is leading to an unprecedented energy consumption crisis in global data centers, which is becoming a significant limitation for the future of AI development [1][4]. Group 1: Energy Consumption Crisis - The competition in the first half of AI development focused on algorithms, while the second half may be constrained by computing power and energy challenges [4]. - The International Energy Agency reports that global data center electricity consumption is expected to surge fivefold by 2030 [4]. - In the "East Data West Computing" initiative, it was observed that for every kilowatt-hour of electricity used for computing, 0.4 kilowatt-hours are used for cooling [4]. Group 2: Liquid Cooling Solutions - Liquid cooling technology presents both advantages and environmental risks; current mainstream liquids, such as certain perfluoroalkyl substances (PFAS), are effective but pose long-term environmental hazards [6]. - New non-PFAS alternatives are emerging in the domestic market, highlighting the need for greener and safer material innovations [6]. Group 3: PUE and Energy Efficiency - The Power Usage Effectiveness (PUE) is a critical metric, with new data centers required to maintain a PUE of ≤1.25, which is seen as a threshold for energy efficiency [8]. - Traditional air-cooled systems typically have a PUE above 1.3, while advanced liquid cooling can reduce it to as low as 1.1 or even approach 1 [8][9]. - A PUE of 2 indicates that half of the electricity is used for computing and half for cooling, with lower values indicating higher energy efficiency [9]. Group 4: Societal Impact of AI - AI has the potential to balance resource distribution, with edge computing enabling high-performance AI models to be accessible in remote areas, enhancing healthcare and education [11]. - AI can help bridge educational gaps caused by geographic disparities, providing equitable access to quality resources [11]. Group 5: Future Considerations - The rapid iteration of technology creates anxiety among tech workers, emphasizing the importance of understanding energy constraints and balancing power with sustainability [13]. - The dialogue underscores the need for technological advancement to align with Earth's capacity, social equity, and human welfare [15].