Huachuang Securities: First coverage of KNOWLEDGE ATLAS (02513) with a "buy" rating leading in independent large models.

Huachuang Securities released a research report stating that KNOWLEDGE ATLAS (02513) is poised to benefit from commercialization acceleration and Token consumption explosion by leveraging Tsinghua technology base and leading MaaS ecosystem in the wave of AI generalization and intelligent body. The company is expected to have operating income of 3.02 billion yuan, 7.108 billion yuan, and 12.278 billion yuan in 2026-2028, with corresponding growth rates of 316.9%, 135.4%, and 72.7%. The first coverage is given a "recommended" rating. The main viewpoints of Huachuang Securities are as follows: Tsinghua technology base establishes first-mover advantage for KNOWLEDGE ATLAS KNOWLEDGE ATLAS originates from the Tsinghua KEG Laboratory, inheriting profound industry-academia-research technological accumulation, and is one of the earliest enterprises in China to deploy large universal models. The company's self-developed GLM pre-training architecture has advantages such as robustness, controllability, and low illusion rate, and is compatible with more than 40 domestic chips, achieving software and hardware co-optimization. KNOWLEDGE ATLAS advocates the core growth concept of "intelligent upper bound Token consumption," reducing unit deployment costs by 50% through model iteration and engineering optimization, and leading the way in the transformation from Vibe Coding to Agentic Engineering. The GLM series will undergo high-frequency iteration in 2025, maintaining its global open source leadership position, driving high-quality Token consumption index growth, and laying a solid foundation for the MaaS platform and developer ecosystem. Top talent drives continuous innovation capability The company has gathered a high-end R&D team led by academicians, with core members having deep accumulation in knowledge engineering, large model pre-training, and data mining. Through internal R&D and parallel cooperation with universities, the company continuously advances cutting-edge technology breakthroughs and establishes a talent training and research transformation system, forming long-term technological reserves in areas such as semantic understanding, AI security, and knowledge graphs, providing continuous impetus for the evolution of large model capabilities. MaaS platform traction accelerates the formation of business ecosystem KNOWLEDGE ATLAS builds a business ecosystem around the MaaS platform, covering cloud-based API, localized deployment, and end-side integration modes. The platform supports a comprehensive model matrix including language, multi-modality, intelligent bodies, and code, allowing customers to quickly call, fine-tune, and customize applications without a professional background. By connecting computing power, devices, institutions, and developers, the company forms an "insight flywheel" to gain scene insights from real deployments and optimize model strategies. As of June 2025, the platform covers approximately 8,000 institutional clients and 80 million devices, making it the most advanced platform in China among independent large model developers in terms of customer reach and commercialization capabilities, promoting the scale landing of AI capabilities in the industrial end. High-speed revenue growth, cloud transformation drives structural optimization The company's revenue has been continuously high-growth from 2022 to 2025, reaching 724 million yuan in 2025, with rapid growth in API and enterprise-level businesses, and significant increases in ARR and call volumes. The business structure is transitioning from local deployment to cloud-based MaaS, with the increased proportion of cloud-based revenue driving long-term profitability improvements, laying the foundation for future economies of scale. AI wave accelerates, KNOWLEDGE ATLAS occupies the core race track Large models are driving AI from specialized to general, with the enterprise market becoming the commercial core, and both cloud-based and local deployments developing in parallel. Industry competition focuses on technology, ecosystem, and business models, with China entering the global leading ranks due to its scene and industrial chain advantages. OpenClaw ignites the intelligent body paradigm, pushing AI from "conversation" to "execution," leading to Token consumption and application ecosystem explosions. With its model and platform advantages, KNOWLEDGE ATLAS is expected to continue to benefit from the new round of industrial leaps. Risk warning: Risks related to technological research and model iteration, uncertainties in commercialization landing and income, risks related to computing power costs and resource supply, intensified industry competition risks, data security, and compliance risks.