“场景定义存储”破局AI时代数据基石难题 产业链共探存储新路径

Group 1 - The core viewpoint of the articles emphasizes the need for a paradigm shift in storage solutions to address the energy consumption challenges posed by the explosive growth of AI computing power [1][2] - The traditional "one-size-fits-all" storage architecture is identified as a key bottleneck for AI development, prompting the proposal of a new paradigm centered around "scene-defined storage" [1] - The CEO of Zhiyu Technology highlighted that as computing power rapidly increases, inefficient data infrastructure leads to significant power and computing waste, positioning storage as the "axis" of AI development [1] Group 2 - Industry experts discussed the necessity for storage solutions in autonomous driving to balance performance, cost, size, and reliability under safety constraints [2] - The focus of supercomputing and intelligent computing centers has shifted from peak traffic to long-term system stability, suggesting a transition from standardized to scene-customized storage systems [2] - The demand for storage devices in industrial control requires a lifespan of 5 to 15 years, necessitating collaboration between suppliers and chip manufacturers from the design phase [2]