机构：华泰证券 研究员：夏路路/侯杰 据百度公司1 月2 日公告，其已根据上市规则PN15 向香港联交所提交有关建议分拆的方案，而香港联 交所已确认可进行建议分拆。我们认为昆仑芯作为国产AI 计算芯片龙头之一，未来其若顺利实现分拆 上市，则上市后更加清晰的信息披露有望助力市场充分认知该业务的内在价值，并推动该业务分部在百 度集团估值中实现进一步的价值重估。展望后续，我们看好在昆仑芯、自动驾驶等子业务线向好发展与 潜在资本市场计划的助推下，百度集团获得相应的价值重估机遇。与此同时，百度集团丰厚的净现金储 备为其估值提供了一定的支撑，后续若公司加大股东回报力度，则有望进一步加强其估值吸引力。建议 持续关注百度各AI 业务项的进展及资本市场计划的推进节奏。 我们认为在AI 时代下，百度集团丰富多元的业务线具备全新发展机遇，后续伴随业务的放量及基本面 的逐步改善，有望进一步推动百度集团的价值重估。1）自动驾驶：12 月22 日，萝卜快跑宣布将于2026 年率先在英国伦敦开启无人驾驶测试及出行服务，是中国无人驾驶首次驶入英国伦敦，进入中国香港以 外的右舵市场（据Wind 报道）。2）AI agent：12 月11 日， ...

Core Insights - Baidu reported Q3 2025 total revenue of 31.2 billion yuan, with core business contributing 24.7 billion yuan, highlighting strong performance in AI-related business which grew over 50% year-on-year [2] - The AI cloud segment saw a 33% year-on-year revenue increase, with high-performance computing subscription revenue surging by 128%, indicating a robust demand for AI infrastructure [2] - AI application revenue exceeded 2.6 billion yuan, driven by subscription models and innovative products like GenFlow 3.0, which has over 20 million active users [2] - AI-native marketing services revenue reached 2.8 billion yuan, a staggering 262% increase, showcasing the effectiveness of AI tools in enhancing productivity and marketing outcomes [3] - Autonomous driving services achieved 3.1 million global orders in Q3, a 212% year-on-year growth, with significant operational milestones in various international markets [3] AI Business Growth - Baidu's AI cloud, applications, and marketing services are experiencing strong growth, with AI cloud revenue increasing by 33% and AI applications surpassing 2.6 billion yuan [2][3] - The launch of the upgraded AI model platform and tools has improved efficiency for enterprise clients, facilitating the development of AI-native applications [2] - The introduction of innovative marketing tools has led to significant improvements in client performance metrics, such as a 60% increase in response speed and a 35% rise in marketing conversion rates [3] Technological Advancements - The release of the Wenxin 5.0 model marks a significant advancement in multi-modal understanding and creative generation capabilities [3] - New hardware products, including the Kunlun M100 and M300 chips, are set to enhance computational power and efficiency, supporting larger-scale model training needs [3] - Baidu's ongoing technological iterations are aimed at building competitive advantages in the AI era, with over 800,000 enterprise clients utilizing AI tools for application development [5]