Core Insights - The company reported a significant improvement in its financial performance for the first half of 2025, achieving a revenue of 988 million yuan, a year-on-year increase of 29.64%, and a net profit of 23.62 million yuan, marking a turnaround from a loss of 6.68 million yuan in the previous year, with a growth rate of 453.67% [1][2][4] Revenue Breakdown - The data traffic business emerged as the core revenue driver, generating 967 million yuan, accounting for 97.93% of total revenue, with a year-on-year growth of 29.56% [2] - The esports gaming sector, while only contributing 2.01% to total revenue, saw a substantial growth of 57.04%, indicating early success in diversifying the business structure [2] Cash Flow and Regional Performance - The net cash flow from operating activities improved significantly to 88.02 million yuan, compared to a negative cash flow of 20.14 million yuan in the same period last year, reflecting a year-on-year growth of 537.05% [2] - Domestic revenue accounted for 98.72% of total income, while overseas revenue, although only 1.28%, grew by 186.84%, showcasing potential for international expansion [2] Business Platforms and AI Integration - The company has established three core business platforms: AI Marketing SaaS, Mobile Application Distribution PaaS, and Spatial Intelligence MaaS, which are strategically aligned to drive growth through AI and data [3][4] - The AI Marketing SaaS platform generated 208.5 million short videos, with AI-generated content accounting for 5.4%, and has been recognized for compliance and technological advancement [3][4] - The Mobile Application Distribution PaaS platform reached a total of 349 million registered users, maintaining stable monthly active users, and supports various applications across multiple sectors [3] Spatial Intelligence and Technological Advancements - The Spatial Intelligence MaaS platform has accumulated over 1.5 million sets of 3D data and 650,000 sets of multimodal data, with significant datasets registered at the Beijing International Big Data Exchange [4] - The company’s strategic focus on data traffic and AI has laid a solid foundation for future growth, particularly in cutting-edge fields such as AIGC and intelligent recommendation systems [4]

8月21日，天娱数科发布2025年半年报。报告期内，公司营业收入9.88亿元，同比增长29.64%，归属于 母公司所有者净利润为2362.01万元，同比上升453.67%。半年报展现的业绩跃升，与AI技术驱动的降本 增效及智能体布局深度绑定，其业务扩张与技术突破在国产芯片适配、智能体生态构建等领域形成独特 竞争力。 随着DeepSeek等国产技术力量的崛起，国产AI正以"软硬协同、场景穿透"的模式重构全球产业格局，在 智能体与芯片两大核心领域形成独特竞争力。这种技术跃迁不仅是单点突破，更是从底层架构到应用生 态的系统性革新。8月21日，DeepSeek正式发布DeepSeek-V3.1，同时支持思考模式与非思考模式。思考 效率更高，相比DeepSeek-R1-0528能在更短时间给出答案。此外，新模型在工具使用与智能体任务中的 表现有较大提升。DeepSeek在官方微信文章标题中称，DeepSeek-V3.1的发布是"迈向Agent（智能体） 时代的第一步"。 据了解，天娱数科在智能体领域的战略布局已形成从底层技术架构到场景落地生态的完整闭环，其自研 的空间智能MaaS平台Behavision，作为通用具身智 ...

Core Viewpoint - The rise of physical AI represents a significant paradigm shift in artificial intelligence, moving from algorithmic interactions in virtual environments to real-world interactions, emphasizing the need for machines to understand and interact with their physical surroundings [1][2] Group 1: Physical AI Development - The company has developed the Behavision platform, which integrates a "perception-reasoning-execution" capability, positioning itself as a leader in the physical AI sector [1] - Physical AI requires a deep understanding of physical properties such as geometric structures, material characteristics, motion laws, and interaction logic [1] - The Behavision platform is focused on creating a key data paradigm called 3D hinge data, designed specifically for training physical AI, which includes understanding the mechanics of objects like doors and drawers [1] Group 2: Data Accumulation and Model Generalization - The company has established a 3D motion capture and multimodal data collection base, accumulating 1.5 million 3D data points and 650,000 multimodal data points, including 200,000 specialized data for robot task training [1][2] - The Behavision platform enhances model generalization through a Real2Sim2Real approach, combining real and simulated data to improve adaptability in complex real-world environments [2] - The platform supports a standardized classification system for robotic bodies and an atomic skill library, enabling quick adaptation across various scenarios such as service, industrial, and logistics [2]

Group 1 - The 2025 Global Humanoid Robot Competition in Shenzhen showcased advanced capabilities such as precise obstacle avoidance, collaborative handling, and dynamic balance, indicating significant technological breakthroughs in the humanoid robotics field [1] - The CEO of Tianyu Digital Science, He Han, proposed a "three-stage theory" for humanoid robot development, outlining a progression from program-controlled demonstrations to imitation based on large models, and finally to an intelligent stage driven by vast 3D data [1] - Tianyu Digital Science is advancing its Behavision platform, which focuses on general embodied intelligence, through its self-developed "Wise Thousand Questions" model and a multi-modal perception hardware system, enhancing its capabilities in perception, reasoning, and execution [1] Group 2 - The company is building a multi-layered, multi-dimensional data ecosystem to accelerate an innovation loop of "technology research and development - data accumulation - scenario validation," with over 1.2 million 3D data points and 500,000 multi-modal data points [2] - The humanoid robotics industry is transitioning from concept to reality, with significant advancements in technology and collaboration among leading enterprises, marking a critical development phase [2] - Tianyu Digital Science is leveraging its "three-stage theory" to establish a solid foundation for future general scene applications, integrating spatial intelligence, 3D data, and cloud-edge computing capabilities to drive the deep integration of embodied intelligence and robotics [2]