比特狂奔，瓦特乏力：AI算力危机与储能的“供血”革命

Core Insights - The article emphasizes that the competition for computing power in the AI era is fundamentally about securing stable and large-scale electricity supply [5][18][19] - It highlights the structural disconnect between the exponential growth of AI computing power and the linear growth of power supply infrastructure, which poses significant challenges for the industry [4][15] Group 1: AI and Power Supply Challenges - AI computing power is experiencing explosive growth, with single-chip power consumption expected to exceed 2 kW and rack power reaching up to 600 kW or more by 2027 [9][10] - The average age of the U.S. power grid exceeds 40 years, leading to slow infrastructure upgrades and challenges in meeting the increasing power demands of AI [3][15] - High volatility in power consumption from AI workloads poses risks to data center stability and the overall power grid [11][12] Group 2: Energy Storage as a Solution - Energy storage is becoming a critical component in the power architecture for AI data centers, transitioning from a backup system to an active component [6][11] - The dual-layer energy storage strategy proposed by NVIDIA includes supercapacitors for rapid response and large lithium batteries for longer energy buffering [12] - The demand for energy storage solutions is expected to rise significantly, with companies like CATL, Huawei, and BYD emerging as key players in the market [21] Group 3: Future Projections and Industry Trends - By 2030, global data center electricity consumption is projected to reach 1500 TWh, with a 160% increase in power demand [14][17] - The article notes that the global AI competition will increasingly focus on breakthroughs in renewable energy, energy storage, and smart grid technologies [19][20] - China's "East Data West Computing" initiative aims to direct computing demands to energy-rich regions, supported by large-scale energy storage facilities [20]