Core Insights - Intel is reportedly negotiating a large-scale wafer foundry contract with Microsoft for the production of chips using the 18A advanced process, attracting interest from tech giants like NVIDIA and Google, challenging TSMC's dominance [1] - The 18A process is described as potentially being an "iPhone moment" for the semiconductor industry, with Intel aiming to position it as an alternative to TSMC's N2 process [1] - Intel's 18A process is expected to enter stable mass production in the second half of this year, with performance claims comparable to TSMC's N2 process [1][2] Company Developments - Intel's new CEO, Pat Gelsinger, is focusing on electronic design automation (EDA), packaging, and wafer foundry services as part of the company's strategic vision [1] - The company is currently in the risk trial production phase for the 18A process, with expectations to transition to mass production within the year [1] Industry Context - The interest from major U.S. tech companies in the 18A process is driven by the need to mitigate geopolitical risks and alleviate tariff pressures, especially in light of potential semiconductor tariffs proposed by Trump [2] - Compared to TSMC and Samsung, Intel's supply chain appears to be more diversified, which could provide a geopolitical advantage if the 18A process achieves stable mass production [2]

"宁波膜智信息科技有限公司"为势银（TrendBank）唯一工商注册实体及收款账户 势银研究： 势银产业研究服务 势银数据： 势银数据产品服务 添加文末微信，加 先进封装 群 作为从事新一代 2.5D/3D堆叠芯片EDA软件设计的研发及产业化企业。珠海硅芯科技有限公司拥 有强大的研发团队，创始人团队从2008年开始研究2.5D/3D堆叠芯片设计方法，是世界最早期研 究前沿芯片架构设计方法的研究团队之一，并在堆叠芯片EDA后端布局、布线、可测试、可靠性等 方面均有世界领先成果。 针对 Chiplet堆叠芯片技术，其自研的3Sheng Integration Platform，支持三维异构集成系统的敏 捷开发与可定制化的协同设计优化，涵盖先进封装设计所需环节的全流程工具。目前，硅芯科技系 列产品已通过先进封装产业验证，完成设计制造闭环，并打造首批客户案例，全方位助力 AI,GPU,CPU,NPU,FGPA,硅光等各类芯片及终端应用领域发展。 直面需求 早先在HiPi联盟大会上， 多位业内顶级设计专家发声Chiplet和3D IC对设计和EDA挑战 。 近年来国内设计三维异构集成芯片的困扰似乎集中于设计出的堆叠结构 ...

Core Viewpoint - Artificial intelligence (AI) is becoming the core engine driving global industrial transformation and technological innovation, with its influence permeating various sectors and redefining chip technology and AI infrastructure [1]. Group 1: AI Industry Acceleration - The AI industry is experiencing comprehensive acceleration, with significant advancements in various sectors [2]. - The "2025 AI Technology Innovation Forum" gathered leading companies and experts to discuss AI trends and innovations [1]. Group 2: EDA Innovations - Shanghai Gaoneng Electronics' Vice President, Ma Yutao, highlighted the challenges and opportunities in analog and custom circuit design in the AI era, emphasizing the dual challenges of precision and speed in EDA tools [3]. - Gaoneng Electronics is focusing on AI/ML integration solutions across the entire chip manufacturing and design optimization process, aiming to transform the semiconductor industry into a data-driven paradigm [3]. Group 3: Flash Memory Demand - Tsinghua Unigroup's market manager, Tian Yue, discussed the rapid growth of the AI server market, predicting it will exceed $233 billion by 2028, with each AI server requiring approximately $100 worth of Flash memory [5]. - Tsinghua Unigroup holds the second-largest market share in SPI NOR Flash, with cumulative shipments exceeding 27 billion units [5]. Group 4: GPGPU Opportunities - Dr. Xiang Tian from Suxian Microelectronics presented on the DeepSeek technology, which significantly reduces computational load during inference, facilitating the deployment of large models on edge devices [8]. - Suxian Microelectronics' "Tianyuan" GPU architecture supports high concurrency and integrates open-source ecosystems, providing comprehensive solutions for AI product deployment [8]. Group 5: AI Power Solutions - Infineon's market manager, Zhou Chengjun, emphasized the need for efficient power applications in AI systems, introducing integrated power modules that reduce power loss to 2% compared to traditional methods [11]. - Infineon is recognized as a global leader in AI power management due to its advanced packaging technology and manufacturing capabilities [11]. Group 6: AI Solutions for SMEs - Zhang Haonan from DeYi Microelectronics discussed the challenges faced by SMEs in deploying AI models and introduced an integrated AI training and inference solution that significantly reduces costs and technical barriers [14]. - The solution supports local processing of large models, offering features like breakpoint training and flexible parameter configuration [14]. Group 7: RISC-V Development - Alibaba's DAMO Academy's Li Jue highlighted the rapid growth of RISC-V, which has surpassed 40% annual growth in mainstream markets, and its potential in high-performance computing and AI acceleration [17]. - The DAMO Academy is iterating on the Xuantie series processors to enhance capabilities for AI applications [17]. Group 8: AI Power Management - AOS Semiconductor's Liu Song discussed the increasing power demands of AI servers and introduced innovative power MOSFET solutions to meet these challenges [19]. - AOS's products are designed to optimize power efficiency and reliability in high-frequency applications [19]. Group 9: Edge AI Trends - Yang Lei from Guangyu Xincheng emphasized the rise of edge AI model chips, which are crucial for the intelligent upgrade of various industries, highlighting the commercial opportunities for hardware companies [21]. - Edge AI offers advantages in real-time processing, reliability, and privacy protection, although challenges remain in performance and cost [21]. Group 10: AI Hardware Evolution - Imagination's Huang Yin discussed the evolution of AI models and the importance of balancing performance with storage and communication needs in edge AI applications [23]. - The future demand for AI hardware will focus on efficiency, integration, and flexibility, necessitating collaboration across the industry [23]. Group 11: High-Performance Thermal Solutions - Lu Cheng from Baidu Technology discussed the increasing demand for high-performance thermal materials, emphasizing the need for customized solutions in various applications [25]. - Current leading thermal materials include aluminum oxide and boron nitride, which meet diverse thermal conductivity requirements [25]. Group 12: Real-Time Fault Monitoring - Wu Jixuan from Texas Instruments highlighted the advantages of edge AI in real-time fault detection systems, showcasing the effectiveness of integrated NPU architectures [28]. - The edge AI solutions provide faster response times and enhanced privacy compared to cloud-based systems [28]. Group 13: Future of AI Chips - A panel discussion featuring Yang Lei, Xiang Tian, and Huang Yin focused on the future design logic and ecosystem collaboration for AI chips, emphasizing the shift from general computing to diverse, modular, and low-power designs [30]. Conclusion - The successful hosting of the "2025 AI Technology Innovation Forum" showcased significant advancements in AI technology across various dimensions, highlighting the importance of industry collaboration and ecosystem synergy [32].