Incident Overview: The company released its 2024 semi-annual report on August 30, 2024. The company's revenue was 64.7653 million yuan, a year-on-year decrease of 43.42%; net profit loss to mother was 0.53 billion yuan; net profit loss due to non-mother was 0.609 billion yuan.

Continued investment in R&D lays an important foundation for long-term development. The company has yet to achieve profits and has accumulated uncompensated losses. There are two main reasons: first, revenue has declined due to adverse factors such as the supply chain; second, the company has continued to invest heavily in R&D to ensure high-quality iteration of smart chip products and basic system software platforms and maintain its technological leadership in a competitive market. In the first half of 2024, the company invested 447.476 million yuan in R&D, accounting for 690.92% of revenue.

However, in the medium to long term, the company's chip products have established cooperation intentions with leading companies in various industries. With excellent evaluation indicators and excellent service reputation of previous products, the company's brand effect gradually highlights the company's focus on Internet communication business development and further promotes the continuous development of independent products, industry solutions, and service businesses centered on the Internet Communications business. The AI chip industry is a highly competitive field, and leading technology is the core foundation for the company to occupy the market and achieve long-term development.

The basic system software platform continues to improve, and the gradual strengthening of the ecosystem will also help the company develop in the long term. 1) Training software platform: The company's training software stack can fully support the distributed training needs of mainstream large models, improve the level of support for each component, and shorten the adaptation cycle of the industry's new models in the company's products. In terms of ecology, the company's training software platform was updated and iterated on PyTorch 2.1/2.3 support, supporting multiple training and inference business scenarios, and implementing a long-term mechanism to quickly follow up on community versions. In terms of large models, the company's training software platform added support for mainstream networks such as Llama3 series and Qwen series, and focused on optimizing parallel training functions. 2) Inference software platform: In terms of AIGC business adaptation, the company's inference software platform follows the application needs and development of the industry, supports and optimizes a series of literary models such as the Llama series and the Qwen series, as well as various mainstream multi-modal models such as StableDiffusion3, to achieve coverage of mainstream AIGC business scenarios such as Wenshengwen, Wensheng map, Wensheng video, and multimodality. In terms of operator optimization, the company developed PyTorch third-party operator acceleration library to achieve specific acceleration of key operators for large models, making the company's inference platform operator calculation efficiency to the leading level in the industry.

Investment advice: The company is a domestic AI chip leader. It actively develops next-generation intelligent processor microarchitectures and instruction sets, continuously optimizes and iterates inference and training software platforms, and promotes the continuous implementation of smart chips and acceleration cards in various key industries such as the Internet, operators, finance, and energy. The company's revenue for 2024-2026 is expected to be 1.536/3.544/5.015 billion yuan, and the corresponding PS for 2024-2026 will be 69X, 30X, and 21X respectively, maintaining the “recommended” rating.

Risk warning: Risk of increased competition in the industry; downstream demand release falls short of expectations.