JP Morgan: Reiterates "overweight" rating on KNOWLEDGE ATLAS (02513) with target price raised to 1800 Hong Kong dollars.

JPMorgan Chase released a research report stating that it has reiterated a "hold" rating on KNOWLEDGE ATLAS (02513) and raised its target price from HKD 1400 to HKD 1800. This target price is based on a 30 times discounted price-to-earnings ratio for 2030. The bank pointed out that with the enhancement of GLM-5.2's capabilities and comprehensive pricing, it has raised its revenue forecast for KNOWLEDGE ATLASAI. The bank has a more positive attitude towards the performance of KNOWLEDGE ATLASAI after the release of the GLM-5.2 model and its pricing update. Compared to GLM-5.1, GLM-5.2 has eliminated lower billing tiers and now charges all usage at higher rates, making the overall pricing higher, with API prices 13% higher than the comprehensive prices of GLM-5.1. As GLM-5.2 still belongs to the GLM-5 model family which has a total of 744 billion parameters and 40 billion active parameters, its performance improvement is mainly driven by reinforcement learning and post-training optimization, allowing the company to achieve higher actual pricing while keeping model costs relatively stable. JPMorgan Chase believes that the large model pricing market is showing a trend of differentiation, with the pricing of mature intelligent products continuing to decline due to the proliferation of capabilities and improvement in reasoning costs, such as DeepSeek being a representative example, continuously driving down market settlement prices and putting pressure on routine and price-sensitive workloads; however, new unlocked frontier capabilities can still command a premium in improving task completion, reducing retries, and opening up high-value application scenarios such as coding, intelligent agents, and automated enterprise workflows. The GLM-5.2 of KNOWLEDGE ATLASAI is a useful case study for testing pricing capabilities. If its API and workflow demand remains resilient after effective pricing adjustments, it will further confirm the company's ability to transform model iterations into high-quality revenue growth with incremental profits. The bank stated that in the short term, attention should be paid to the demand resilience of GLM-5.2, the model iteration progress of competitors like Kimi K3 and DeepSeek V4.1, as well as the upcoming release of the GLM-5.5 model by KNOWLEDGE ATLASAI expected in August.