AI行业跟踪报告之第49期：存算一体产业趋势渐起，建议关注恒烁股份炬芯科技

Investment Rating - The report maintains a "Buy" rating for the industry, specifically highlighting companies such as Hengshuo Co., Juchip Technology, and others [13]. Core Insights - The emerging trend of in-memory computing technology significantly reduces data transmission latency and power consumption, enhancing overall system performance [3][10]. - The report emphasizes the importance of in-memory computing in addressing the "von Neumann bottleneck," which has become a critical issue in the era of big data and artificial intelligence [4][19]. - The report identifies three main types of in-memory computing technologies: Processing Near Memory (PNM), Processing In Memory (PIM), and Computing In Memory (CIM), each with distinct advantages and applications [50][12]. Summary by Sections Industry Overview - The in-memory computing technology is positioned as a solution to the limitations posed by traditional computing architectures, particularly in high-performance computing and energy efficiency [19][26]. - The report notes that the slowdown of Moore's Law has necessitated the exploration of new computing paradigms, with in-memory computing providing a viable alternative by minimizing data movement [8][18]. Key Technologies - PNM technology reduces data transmission distance, suitable for high-frequency data exchange scenarios [11]. - CIM technology integrates computation and storage in the same physical location, offering significant advantages in parallel computing and deep learning tasks [12]. - The report discusses the advantages of various storage technologies, including SRAM, DRAM, and emerging non-volatile memory technologies like ReRAM and MRAM [53][25]. Market Trends - The report highlights the growing demand for in-memory computing in sectors such as artificial intelligence, edge computing, and data centers, driven by the need for high performance and low power consumption [54][19]. - Companies like Hengshuo Co. and Juchip Technology are noted for their advancements in in-memory computing technologies, with specific applications in AI and IoT devices [40][41]. Company Profiles - Hengshuo Co. focuses on the development and sales of storage chips and MCU chips, with ongoing projects in in-memory computing technology [41][66]. - Juchip Technology has adopted a mixed-mode SRAM-based CIM approach, achieving high energy efficiency and performance in AI applications [77][81]. - The report mentions ByteDance's investment in Xinyuan Semiconductor, which specializes in ReRAM technology, indicating a strategic move towards enhancing in-memory computing capabilities [63][36].