Haitong Securities released a research report saying that the storage industry has ushered in significant growth on the demand side, driven by the rapid development of AI technology, and the supply side has gradually improved the supply and demand structure through the production reduction measures of NAND Flash manufacturers, and the industry is expected to usher in a recovery. The rapid development of AI mobile phones, AI PCs, AIoT, and smart cars has further increased the capacity of memory units. BIWIN Storage (688525.SH) has entered high value-added fields such as AI glasses through the integration of layout storage + advanced packaging and testing, and has great potential for future development.

The main views of Haitong Securities are as follows:

Optimistic about the benefits of the storage industry

(1) On the demand side: DeepSeek accelerates the full flowering of AI device-side applications, and AI+ semiconductors drive the increase in storage demand; (2) Supply side: The original NAND Flash manufacturers restarted production reduction, laying the foundation for the rebound of storage prices in the second half of the year. The bank believes that memory module companies will significantly benefit from the improvement of supply and demand structure and the expectation of price recovery, especially high-quality companies that can cooperate with advanced packaging and testing and AI device side. BIWIN Storage has integrated storage + advanced packaging and testing, invested in wafer-level advanced packaging and testing manufacturing, and covered well-known overseas AIoT customers, cutting into the field of high ASP of AI glasses.

The far-reaching impact of DeepSeek is mainly reflected in two dimensions

1) The low-cost solution provided by DeepSeek will provide new training ideas for the industry under the same training effect, and large manufacturers will be deeply inspired by it when training large models, and at the same time, domestic cloud vendors will also shift their attention to the potential development of domestic computing power cards; 2) The open source of DeepSeek will bring global AI equality, and the motivation to deploy open source models locally at both ends of B/C will be sharply enhanced, in other words, the demand for computing power cards for cloud inference and the continuous increase of local computing and storage to support large models will become a medium and long-term industry trend.

The bank believes that the application side is expected to benefit deeply, and the storage demand will be comprehensively improved with the implementation of AI+ applications

　　1) AI mobile phones: DeepSeek's cost reduction and open source is expected to push up the configuration of device-side computing and storage, and while cloud computing power dominates, device-side collaboration is expected to usher in a significant enhancement. In addition, AI mobile phone brand manufacturers are expected to gradually increase their voice in the context of continuous data integration with third-party APP manufacturers. 2) AI PC: In the early stage, the penetration of AI PCs in the B-end and C-end is limited by data security and high consumption thresholds, DeepSeek can help B-end private clouds deploy local large models to achieve comprehensive privacy protection, reduce the need for data transmission by processing data on local devices, thereby reducing the risk of data leakage, and C-end users will also strengthen their willingness to deploy locally, and the growth logic of AI PCs is expected to be gradually reshaped. (3) AIoT: End-to-end applications are expected to blossom in an all-round way, and the penetration of AR glasses is expected to be further enhanced. (4) Automotive intelligence: The bank believes that smart cars are expected to introduce open source large models to become larger AIoT terminals, and AI may deeply empower the whole life cycle of automobiles. In 2024, the average capacity of DRAM and NAND Flash in various AI extension applications, such as smartphones, servers, and laptops, will grow, and the bank believes that in 2025, with the further increase in the penetration of AI applications, the number of storage units will continue to grow, and storage prices are expected to recover under the reduction of storage production + improved demand.

At the beginning of 2025, all NAND Flash OEMs will take more resolute measures to reduce production and reduce the annual production scale, hoping to effectively reduce the growth rate of supply bits, mainly by reducing the utilization rate in 2025 and delaying the process upgrade, which is conducive to quickly reducing the pressure of market supply and demand imbalance and laying the foundation for price rebound. The 25Q1 NAND Flash market continues to face the challenge of oversupply, resulting in a continuous decline in prices and losses for suppliers. The bank believes that the supply and demand structure of the NAND Flash market is expected to improve significantly in the second half of the year, including the original factory production cut, smartphone inventory depletion, AI and DeepSeek effect and other factors will push up the demand for NAND Flash, thereby alleviating the oversupply situation, and it is expected that the price will also recover in the second half of the year. In terms of DRAM, the bank believes that DRAM prices are also expected to recover in the second half of the year as demand for networking improves.

BIWIN's integrated layout of storage and storage + advanced packaging and testing, and fixed investment in wafer-level advanced packaging and testing manufacturing

Focusing on the semiconductor memory industry chain, BIWIN Storage has built a business model of "integration of R&D, packaging and testing". The total amount of funds raised by the company through the issuance of shares to specific targets does not exceed 1.9 billion yuan, which will be used for the expansion and construction project of Huizhou BIWIN's advanced packaging and testing and memory manufacturing base, and the wafer-level advanced packaging and testing manufacturing project. With its subsidiary Huizhou BIWIN as the advanced packaging and testing and memory manufacturing base, Huizhou BIWIN's packaging process is leading in China, and currently has the mass production capacity of advanced processes such as 16-layer stacked die, 30~40μm ultra-thin die, and multi-chip heterogeneous integration, reaching the international first-class level. BIWIN has provided foundry services to wafer manufacturers, IC design companies, and memory manufacturers, forming a new business growth point.

BIWIN Storage covers well-known overseas AIoT customers and enters the field of high ASP of AI glasses

BIWIN's ePOP series products have been applied to smart watches, VR glasses and other smart wearable devices by well-known companies such as Google, Meta, and Xiaotiancai. The company's eMCP and uMCP series products have been widely recognized by customers of smart phones and tablets. According to the Weishen Wellsenn XR WeChat public account, in the BOM composition of Ray Ban Meta smart glasses, the hardware cost of ROM+RAM is 11 US dollars, accounting for 12.35% of the motherboard chip (89.1 US dollars), and the memory module eMCP is a field with a high proportion of semiconductor value in AI glasses.

Risk warning: the recovery of terminal demand is less than expected; The penetration of AI applications is not as good as expected; The price recovery of storage terminals is less than expected; The process of domestic substitution of semiconductors is less than expected.