人工智能时代，EDA巨变

Core Viewpoint - The semiconductor industry is undergoing a significant transformation driven by AI, which is leading to a shift from traditional SoC designs to Chiplet architectures, enhancing performance and flexibility in chip design [2][5][9]. Group 1: Chip Design Evolution - The traditional SoC design is becoming inadequate due to increasing demands for AI processing power and the challenges posed by advanced manufacturing processes [5][6]. - The Chiplet model allows for the separation of different functional modules, improving yield and flexibility while enabling heterogeneous integration of various manufacturing processes [5][6]. - The transition from SoC to CoWoS (Chip on Wafer) can enhance bandwidth by 4 to 8 times, and moving towards SoW (System on Wafer) can increase scale by up to 40 times [6]. Group 2: EDA's Role in AI - EDA (Electronic Design Automation) must evolve to support the new demands of AI, requiring a shift from chip-level considerations to system-level integration, including power, thermal management, and interconnects [7][9]. - The introduction of AI into EDA processes aims to transition from rule-driven design to data-driven design, significantly improving design efficiency [9][10]. - The company is focusing on a dual strategy of "EDA FOR AI" and "AI+EDA," leveraging its expertise in Chiplet and system design to support AI chip development [9][10]. Group 3: Future Directions - The integration of AI into EDA tools is expected to lead to continuous learning and iteration, transforming EDA from a passive tool to an active collaborator in the design process [12]. - The next phase of AI development is anticipated to be the era of physical AI, where multi-physical simulation technologies will play a crucial role [12].