Incidents:

On April 26, the company released its 2024 quarterly report. In 24Q1, it achieved revenue of 543 million yuan, +64.89% year-on-year, and net profit of 68 million yuan. Gross margin increased 1.2 percentage points year over year and 1.7 percentage points month over month.

The downstream market picked up, and automotive electronics, industrial applications, and machine vision achieved relatively rapid growth:

According to the announcement, downstream demand recovered in the first quarter, and terminal customer inventories returned to normal. The company achieved rapid growth in the fields of automotive electronics, industrial applications, machine vision, etc., and the number of mass production customers and projects for the flagship RK3588 chip continued to increase, entering a period of rapid growth. The trend of localization and miniaturization of AI models at the edge and terminals is more clear, which is expected to accelerate the development of various AIoT industries that the company has been deeply involved in for a long time. The next new products, such as the RK356X and RV1106/1103, are also being released at an accelerated pace. New products such as the next-generation mid-range and high-end AIoT processor RK3576 and the AI audio processor RK2118 have laid a solid foundation for the company's growth in 2024.

It is expected to benefit long-term demand for end-side generative AI computing power improvement:

With the rapid increase in the parameter scale of large models, the parameter scale has entered the 100 billion and trillion era. The general large model has excellent performance, but it also brings greater hardware investment and power consumption. Recently, leading technology companies such as Qualcomm, Apple, Intel, and AMD are all speeding up the deployment of generative AI and exploring applications in mobile phones, PCs, and other terminals. Qualcomm proposed in the “Hybrid AI White Paper” that hybrid AI is a trend of large-scale AI development, that is, model training is implemented in the cloud; according to model complexity, the inference work is placed on the terminal side. AI inference on the terminal side can bring cost, energy consumption, performance, privacy, security and personalization advantages, and help realize intelligent computing anytime, anywhere. According to the company's annual report, the company continues to strengthen chip support for AI applications. First, it continues to upgrade NPU IP and related tool chains to improve the efficiency of supporting mainstream model architectures represented by the Transformer architecture to help deploy models on the edge side and end side; the second is to roll out AI algorithms such as vision, video, and audio, and work with upstream and downstream partners to develop AIoT products.

Investment advice:

We expect the company's revenue from 2024 to 2026 to 2026 to be 2,688 million yuan, 3.469 billion yuan, and 4.44 billion yuan, respectively, and net profit to mother of 313 million yuan, 597 million yuan, and 853 million yuan respectively.

Considering the company's high R&D investment and the impact of the semiconductor downturn cycle, the current profit may not fully reflect the company's long-term profitability. Using the PS valuation method, the company is expected to benefit from the long-term increase in demand for generative AI. According to Wind's unanimous expectations, Quanzhi Technology, Hengxuan Technology, and Haiguang Information's average PS value of 11.5 times was selected as a reference, corresponding to a target price of 73.4 yuan to maintain a “buy-A” investment rating.

Risk warning: Market demand falls short of expectations; new product launches fall short of expectations.