CITIC Securities released a research report stating that KNOWLEDGE ATLAS (02513) is a leading enterprise in China's general large model sector. Over the past two years, focusing on the internet and technology enterprises as its core market, the company has achieved sustained revenue growth exceeding 100%. The domestic large language model market is projected to grow twentyfold over the next six years, with enterprise-level demand poised to dominate a market opportunity worth hundreds of billions. The company holds a competitive advantage in this relevant market. The report is optimistic that application deployment will continue to drive rapid growth in model demand, and that the company will leverage its superior model capabilities to further penetrate the enterprise market, sustaining high revenue growth. The report assigns the company a 2026 price-to-sales (PS) ratio of 30 times, corresponding to a target market capitalization of HK$53.9 billion.

Company Overview: A leading domestic general AI model enterprise, possessing top-tier scientist resources. The company is dedicated to developing advanced general large models and is China's largest independent general large model developer, serving over 8,000 customers with a global developer community exceeding 45 million downloads. Its core management team boasts profound technical backgrounds, and the R&D team includes top scientific resources, such as Academician Zhang Bo. In 2024, the company's revenue reached RMB 312 million, a year-on-year increase of 150.9%. For the first half of 2025, revenue was RMB 191 million, surging 325.0% year-on-year. R&D expenses for H1 2025 totaled RMB 1.595 billion, of which computing power investment accounted for RMB 1.145 billion. The gross margin for its on-premise deployment business remains above 60%. Supported by its model performance, the scale effects of the company's operations are expected to become increasingly evident.

Industry Overview: The large language model market is anticipated to expand twentyfold over the next six years, with enterprise-level applications dominating a massive opportunity. The company holds an advantageous position in the enterprise market. According to Frost & Sullivan forecasts (cited in the company's prospectus), the size of China's large language model market reached RMB 5.3 billion in 2024 and is projected to grow to RMB 101.1 billion by 2030, representing a CAGR of 63.7% from 2024 to 2030. Frost & Sullivan estimates that China's enterprise-level large language model market will reach RMB 90.4 billion by 2030, accounting for nearly 90% of the total market. Based on Frost & Sullivan data and expert interviews (cited in the prospectus), KNOWLEDGE ATLAS held a 6.6% market share by large language model revenue in 2024, making it the largest independent large language model provider.

Model Capabilities: High cost-performance, low hallucination rate, comprehensively covering diverse enterprise needs, with excellent coding and Agent capabilities in GLM-4.7. The company's models cover a parameter scale ranging from on-device deployment (9B) to flagship models (355B), meeting the usage requirements of different enterprises. Its models are characterized by high cost-effectiveness and low hallucination rates. According to Stanford University's "2025 AI Index Report," test results from the Hughes Hallucination Evaluation Model (HHEM-2.1-Open) showed that the company's GLM-4-9B model achieved one of the lowest hallucination rates (1.3%) among top-tier models. The latest flagship model performs exceptionally well in Coding and Agent capabilities. The GLM-4.7 model, released in December 2025, set a new State-of-the-Art (SOTA) benchmark for domestic models in coding. Based on test set performances, the model has reached international advanced levels in areas such as logical reasoning, code generation, and screen recognition, narrowing the gap with leading overseas models like Gemini 3.0, and has gained recognition from enterprises on overseas platforms like OpenRouter and Fireworks.

Growth Analysis: New products are driving a gradual increase in the proportion of cloud deployment, while the on-premise deployment business is expanding into new industries. Currently, on-premise deployment contributes 85% of revenue, while cloud-based revenue has grown from 0% to 15% over the past two years; the report anticipates the cloud revenue share will continue to rise in the medium term. 1) On-premise deployment business: The average annual fee per customer exceeded RMB 2 million in 2024, having doubled over the past two years. By industry, 50% of 2024 revenue came from the internet and technology sectors. The company plans to prioritize the consumer electronics and IoT industries in the next six months, and medium-term strategy involves leveraging model advantages to enter high-barrier industries like education and healthcare. The report believes that as enhanced model capabilities accelerate application deployment across various fields, the on-premise business is poised for rapid simultaneous growth in both customer numbers and average spending per customer. 2) Cloud business: New products like GLM CodingPlan and AutoGLM demonstrate excellent performance. The latest GLM-4.7 ranked first among open-source models on Code Arena and LiveCodeBeach, driving a rapid increase in global paying users in 2025. Referencing the company's official WeChat account, the number of paying customers on its open platform, Bigmodel.cn, increased tenfold for the full year, and high-performance code tools are expected to have a greater impact on cloud revenue by 2026.