CICC: Assigns "outperform" rating to KNOWLEDGE ATLAS (02513) with a target price of HK$688.

Zhongjin released a research report stating that the revenue of Knowledge Atlas (02513) is expected to be 7.2/17.5/36.7/76.7 billion yuan from 2025 to 2028, with a CAGR of 120%. The bank is optimistic about the company's ability to iterate on the base model and realize the coding scene, and values the company based on the valuation of overseas leading large model companies, giving the company an outperform industry rating and a target price of 688 Hong Kong dollars (40 times P/S in 2028, discounted at 7% in the future). The company is currently trading at 25 times P/S in 2028, with a 42% upside potential from the current stock price. The company is a leading Chinese and world-class large model manufacturer. Zhongjin's main points are as follows: Originating from Tsinghua University's technological achievements, the technical foundation and model capabilities are the foundation. The company was established in 2019, focusing on in-depth development of the GLM series base model and polishing the core capabilities of models such as Coding, Inference, and Agentic. The latest generation base model GLM-5 has reached the SOTA level in multiple benchmark scores such as HLE and SWE, and has received extensive praise from users at home and abroad. Model commercial value realization, empowering various industries. The company relies on the MaaS platform to output model capabilities, and the bank expects a CAGR of over 130% in revenue from 2023 to 2027. API revenue will become the main growth engine. The bank estimates that the company's API-related ARR will be close to 600 million yuan by early 2026, which is tens of times higher than the same period last year. At the same time, the company covers various industries such as the internet, software, and chips, focusing on releasing the value of models to empower customers to improve productivity. The AI Coding TAM could reach trillions of yuan in size, and Knowledge Atlas has a leading advantage. AI Coding, as the fastest landing scenario for AI, is still in the early stages of penetration. According to the bank's estimates, TAM will reach trillions of yuan in size. Knowledge Atlas focuses on refining the Coding scene, with core advantages such as low hallucination rate, high stability, strong reasoning and tool usage capabilities, and is expected to continue leading in the Coding scene and extend to more enterprise scenarios. The bank's biggest difference with the market? The bank is optimistic about Knowledge Atlas's ability to output base model capabilities and continue to lead in the AI Coding scene, potentially covering more domestic and foreign customers. Potential catalysts Release of a new generation of models; high growth in API and Coding ARR. Risks R&D progress is slower than expected; commercial expansion is slower than expected; risks related to intellectual property rights; risks of funding shortages; risks of intensified competition; risks of data security.