The Zhitong Finance App learned that Guojin Securities released a research report saying that the improvement in the smartphone decline in the first half of 2023 in the second half of the year meant that consumer pessimism about mobile phones began to ease. At the same time, the increase in the average price of mobile phones meant that consumer acceptance of high-end smartphones increased. As AI phones provide a new user experience and consumers are willing to try them out early, AI phones will drive further recovery in the market. With AI phones entering the market in '24, memory will usher in a more energetic capacity increase cycle. Judging from 23Q4, the rate of memory growth has already begun to increase. It is recommended to focus on feature machine replacement opportunities in overseas markets, especially in the Middle East & Africa, as well as the AI mobile phone storage and main chip industry chain.

Guojin Securities's views are as follows:

Global smartphone sales began to recover in the second half of 2023, and the market environment improved.

According to IDC data, global smartphone sales in 2023 were about 1,167 billion, down 3.2% year on year. 23Q1, Q2, Q3, and Q4 sales volume was approximately 2.68, 2.68, 3.04, and 324 million units, -14.6%, -6.82%, +0.33%, and +7.8% year-on-year.

The bank believes that the improvement in smartphone decline in the first half of 2023 means that consumer pessimism about mobile phones has begun to ease. At the same time, the increase in the average price of mobile phones means that consumer acceptance of high-end smartphones has increased. As AI phones provide a new user experience and consumers are willing to try them out early, AI phones will drive further recovery in the market. The Middle East & Africa are in the process of switching from feature phones to smartphones. In 2023, the Middle East & Africa region sold about 244 million units of mobile phones, 104 million units of smart phones, and 140 million units of smart phones. Smartphones accounted for about 57.4%, up about 9.5 percentage points from the lowest point in 2017. There is still plenty of room for feature phones to switch to the smartphone market in the future.

Consumers are looking forward to AI phones, and at the same time, AI has begun to change the shape of mobile phones.

The number of pre-orders for the Galaxy S24 series in Korea over a period of one week reached 1.21 million units, breaking the sales record for Samsung phones. Currently, mobile applications related to AI are still limited to chatbots, simple schedule management, smart notes, etc., but AI has the potential to empower existing image and video apps and break through productivity limitations. In the future, mobile phones can complete LAM (LargeActionModel) training combined with large language models on the terminal side to simplify the operation imagined by RabbitR1 while maintaining strong social attributes.

All major model-side manufacturers have begun to deploy LLM, and have accumulated relevant experience on the AI phone model side, paving the way for their own AI phones.

On February 21, Google released the Gemma open source model, which is divided into two versions, 2B and 7B. Each includes a pre-trained basic version and an instruction optimized version. Compared to Llama2, this big open source model performs better in terms of reasoning, data, programming, etc. At the same time, models with 2B and 7B parameters are also ideal for training on mobile phones, PCs, and mobile devices. They do not require quantitative data processing, and have a processing capacity of up to 8Ktokens. In the field of image generation models, Samsung has its own Gauss model that can generate and edit creative images, including changing and adding style. Xiaomi has developed its own large image model and implemented functions such as AI image expansion & removal. Apple collaborated with researchers at the University of California to create MGIE, which can handle all kinds of editing situations, from simple color adjustments to complex object manipulation. The model can also perform global and partial edits according to the user's preferences.

As AI begins to be deployed on the end side in the second half of 2023, there is no time to delay the increase in memory for AI phones.

The bank believes that with AI phones entering the market in '24, memory will usher in a more energetic capacity increase cycle. Judging from 23Q4, the rate of memory growth has already begun to increase. At the same time, the demand for memory bandwidth etc. of AI phones has increased, and the bank believes that the penetration rate of LPDDR5 will accelerate. The SoC architecture has changed, the status of NPUs in mobile phone SoCs is constantly improving, and customized CPUs, GPUs, and NPU cores are also more important. Qualcomm and MediaTek have changed the AI mobile chip architecture one after another, from reducing the number of energy efficiency cores to increasing the number of performance cores to enhance the chip's performance in computing. At the same time, Qualcomm is about to abandon the official ARM CPU and switch to self-developed NuviaCPUs to further improve chip performance. The Apple A17Pro focuses on the NPU side, doubling the computing power compared to the A16.

According to IDC data, the global sales volume of folding screen phones in 2023 was about 18 million units, up 28% year on year. At the same time, the price dropped steadily from 1,800 US dollars/unit 2 years ago to around 1,200 US dollars/unit. As the performance and appearance of folding screen phones continue to improve and prices drop, the bank believes that sales of folding screen phones are expected to increase further in 2024, and Hinge, MIM, and OLED industry chain companies will also benefit from this.

Investment advice and valuation

It is recommended to focus on functional machine replacement opportunities in overseas markets, especially in the Middle East & Africa; it is recommended to focus on folding screen mobile phone related industry chain companies: Ruisheng Technology (02018), Dongmu Co., Ltd. (600114.SH), Jingyan Technology (300709.SZ), Tonglian Precision (688210.SH), Xinwei Communications (300136.SZ), etc.

Risk warning: smartphone sales fall short of expectations; end-side AI model development speed falls short of expectations; hardware updates fall short of expectations; geopolitical risks;