Incident: The company released the 2024 semi-annual report. 2024H1 achieved total operating income of 2.978 billion yuan, yoy +18.40%, net profit to mother -0.192 billion yuan, and net profit after deducting non-return to mother of 0.004 billion yuan, in line with market expectations.

Single Q2 shows a recovery in profitability. Looking at 2024 Q2, the company achieved total operating income of 1.885 billion yuan, yoy +32.07%, net profit attributable to mother of -0.009 billion yuan, and net profit of 0.037 billion yuan after deducting unexplained losses. In terms of gross margin, the overall gross margin of the company in 2024 Q2 was 14.75%, a significant improvement compared with 11.81% in the 2023 single Q2.

Large model capabilities continue to iterate to form a “family bucket” of product matrices. The company released the Jiadu Zhixing Transportation Model V2.0 in June 2024. This version covers different parameter scales of 1 billion to 100 billion, forming a “family bucket” of product matrices such as PCITransgpt, the large visual model Trancore-V, the multi-modal large model Trancore-M, and the large spatio-temporal decision model Trancore-TS. The Jiadu Zhixing Transportation Model V2.0 not only solved the underlying technical problems of the big model, but also explored and achieved excellent results at the application level, achieving an upgraded transformation from technical best practices to business value tools.

The official release of the Jiadu Zhixing Transportation Model V2.0 will deeply empower various business scenarios such as intelligent urban rail customer service, intelligent operation and maintenance, emergency management, safe operation, passenger service, passenger flow prediction, energy saving and environmental control, and help urban rail digital intelligence achieve faster technology iteration and more efficient application implementation.

AI transformation capabilities are improved, and application scenarios are implemented quickly. 1) In terms of products, the company has officially launched a car driving management e-book product based on the cooperation between the Jiadu Zhi Xing Da Model and the Guangzhou Traffic Police Detachment, with an accuracy rate of more than 95%. The company also incubates large model products covering the two core traffic management businesses of command and law enforcement. These three types of business scenario products have all received positive feedback from the Guangzhou Traffic Police Detachment and will continue to be updated and iterated. The company's self-designed and developed comprehensive product line solution for traffic management intelligent sensing+information control has been initially implemented at 0-1. The solution includes the lightweight design and development of a holographic series of products. Currently, the traffic lights developed by the company have passed the safety review by the Guangzhou Traffic Police Detachment and obtained the opportunity for pilot implementation. 2) In terms of hardware, the company and Huawei jointly released the “Zhixing Transportation Model Training and Promotion All-in-One” product series to build integrated solutions for users and promote the privatization and delivery of large-scale models in the industry.

Maintain an “Overweight” rating. Considering that the company continues to consolidate its technical base around the two major scenarios of rail transit and transportation, related AI technology is expected to continue to empower the company's existing business scenarios. We expect that in 2024-2026, the company's net profit to mother will be 0.083/0.193/0.309 billion yuan, respectively, corresponding to the closing price on September 13, 2024. PE will be 97.8/42.2/26.3 times, respectively, maintaining the “gain” rating.

Risk warning: The economic downturn exceeded expectations, fiscal spending fell short of expectations, and international competition intensified.