Summary of Moore Thread (688795.SS) Management Call Company Overview - **Company**: Moore Thread (688795.SS) - **Industry**: Semiconductor, specifically focusing on General GPU technology - **Recent Developments**: Introduction of new generation products including HuaShan and LuShan GPUs based on the HuaGang architecture [3][4] Key Points Discussed 1. General GPU Strengths - Moore Thread is focusing on developing General GPUs that offer comprehensive features beyond AI computing acceleration, such as 3D graphics rendering and UHD video encoding/decoding [4] - The General GPU is designed to support large-scale AI training and is compatible with high technology entry barriers [4] 2. AI Computing Platform Applications - The company is optimistic about the rising demand for AI chips in the Chinese market, driven by the accelerated iteration of new AI models and applications [1] - Management highlighted the importance of their General GPU in supporting multi-modal AI models that require higher computing power for tasks like video and 3D content generation [1][4] 3. Customer Base and Future Expansion - Moore Thread is expanding its client base across various sectors including AI data centers, AI edge devices, and gaming graphics [9] - The recent launch of the S5000 GPU aims to enhance performance for data center clients, with ongoing product verification with local CSP clients [9] - Management is positive about the company's ability to adapt its platform to local AI models, such as DeepSeek, and assist clients in AI deployment [9] 4. Full-Stack Solutions - The company offers a full-stack solution that includes chips, OAM, and computing clusters, along with an in-house software stack [5] - This comprehensive approach is aimed at driving the implementation of AI across various applications, including entertainment, smart medical, and industrial sectors [8] 5. Market Trends and Client Demand - There is a noted increase in demand for enhanced AI models and applications from local suppliers, necessitating platforms with higher computing power and efficiency [2] - Management's positive outlook on the performance of local GPUs aligns with the broader trend of advanced node expansion in the semiconductor industry [2] 6. Demonstrations and Developer Support - During the MUSA developer conference, Moore Thread showcased over 80 demo applications powered by its computing platform, emphasizing low learning costs for developers transitioning to their platform [8] Additional Insights - The management's discussions reflect a strong commitment to innovation in GPU technology and a proactive approach to meeting the evolving demands of the AI market in China [1][2][4] - The strategic focus on multi-modal applications and comprehensive product offerings positions Moore Thread favorably within the competitive landscape of AI and semiconductor industries [5][9]

Investment Rating - The report assigns a "Buy" rating to Kingsoft Office, Kingdee, and Empyrean [5][31]. Core Insights - The momentum of AI-native applications and software with AI features remains strong, particularly in the areas of agentic AI and multi-modal AI models [1][4]. - AI agents are expected to become the new user interface for enterprises, enhancing productivity through proactive responses to environmental changes [4][12]. - The release of upgraded multi-modal AI models focuses on generating and editing various content types with improved quality and lower costs [4][13]. - There is a solid project pipeline for enterprise application wins, particularly in AI model deployment, indicating a larger scale of AI projects compared to traditional ERP or system upgrades [21][4]. Summary by Sections AI Agents and Applications - AI agents are being adopted by enterprises to complete tasks independently, with companies like Manus launching general AI agents and Kingdee introducing multiple specialized AI agents [4][12]. - The report highlights the potential of AI agents to improve user experiences in various sectors, including finance and travel [4][12]. Multi-modal AI Models - Recent upgrades in multi-modal AI models have been made by vendors, focusing on high-quality content generation across different media types [4][13]. - Companies like Stepfun and Wondershare are developing advanced tools for image and video editing, enhancing user capabilities [4][13]. Software Project Wins - The report reviews enterprise application project wins, noting a solid momentum in AI model deployments from late April to the present [21][4]. - The scale of AI projects is generally larger due to the inclusion of integrated solutions, which often require higher computing hardware costs [21][4]. EDA and IP Software Expansion - Local EDA suppliers are accelerating product launches to capture localization opportunities, with new tools being introduced for mixed-signal SoC and digital simulation [4][21].