Embrace the AI ​​era and win the future of the storage industry! The 3rd GMIF2024 Innovation Summit was successfully held in Shenzhen

In the golden autumn of Shenzhen, let's join in the grand event. On September 27, the "3rd GMIF2024 Innovation Summit (Global Memory Innovation Forum)" was grandly held in Shenzhen, hosted by the Semiconductor Investment Alliance and Shenzhen Memory Industry Association, co-organized by the Guangdong Integrated Circuit Industry Association and Shenzhen Semiconductor Industry Association, and undertaken by Aiji Micro Consulting (Xiamen) Co., Ltd. and Haitong Securities Co., Ltd. The theme of this summit is "AI-driven storage recovery". It brings together mainstream terminal manufacturers, module manufacturers, packaging and testing manufacturers, equipment and material manufacturers and other leading companies and investment institutions in the storage industry chain to jointly discuss hot topics such as industry product innovation, technological evolution, and coordinated development of the industry chain, actively promote the coordinated development of the industry, and win the future together.

AI applications are accelerating the implementation of the storage industry, opening up new opportunities

Currently, the explosive development of AI is redefining the global semiconductor industry landscape. The widespread application of AI has driven the demand for high-performance, high-computing power chips, and has also brought huge growth space for high-bandwidth, large-capacity, and low-power storage chips.

At the beginning of the summit, Sun Rixin, president of the Shenzhen Memory Industry Association, said in his speech that in recent years, with the rapid development of technologies such as artificial intelligence, 5G, big data and the Internet, the global storage industry is entering a new period of change . President Sun said that the purpose of the GMIF Innovation Summit is not only to focus on the memory itself, but also to focus on the entire upstream and downstream value chain of the memory, covering storage media, solutions, system platforms, test equipment and other key areas. We firmly believe that only through the collaborative cooperation of the entire industry chain can we continue to promote industry progress and stay ahead in global competition.

Zhang Xiaofei, chief electronics industry analyst at Haitong Securities, said that from the current memory market, HBM3e affects DDR5 production, and DRAM prices have limited declines; in terms of NAND, server terminal inventory adjustments are coming to an end, and AI has driven demand for large-capacity storage products, driving prices to continue to rise in Q2. The recent NAND growth has narrowed, but it is still better than market concerns. Looking ahead, AI terminal application penetration is accelerating, and computing storage demand continues to rise.

Cai Yimao, dean of the School of Integrated Circuits at Peking University, said that semiconductor memory is the largest branch of the integrated circuit industry. Entering the post-Moore and artificial intelligence era, storage technology faces major challenges. On the one hand, the integration density and reliability of traditional memory below 28nm are severely restricted, and breakthroughs in underlying technology are urgently needed. On the other hand, the demand for AI computing power is growing rapidly, and traditional computing chips have high hardware overhead and high energy consumption, which cannot meet the high energy efficiency requirements of smart devices. It can be said that breakthroughs in underlying unit and process integration are the core of memory technology development, and new storage technology forms are becoming more mature.

"AI applications are ubiquitous and have entered our daily lives in various forms," ​​said Prasad Alluri, vice president and general manager of Micron Technology's client business unit. "For example, smartphone manufacturers have introduced AI features in high-end phones and have increased the memory capacity of LPDDR 5 to between 12 gigabytes and 16 gigabytes to accommodate the ever-increasing data sets. In addition, storage technology has been iterated to UFS 4.0, which has more than doubled the power efficiency compared to the previous generation. In terms of edge devices, AI will increase autonomous driving from level 0 to level 5, and industry insiders expect that the bit density required for automotive memory will increase by about 30 times, while the non-collision bits within the power supply will increase by nearly 100 times. From data centers to edge devices, and all the enhancements in memory and storage environments, would not be possible without our improvements to generation after generation of technology."

Li Yan, vice president of Western Digital's advanced flash memory technology, pointed out that 3D NAND flash memory is a highly competitive and constantly improving market. The number of stacks has increased rapidly, from 48 layers in 2017 to 64 and 96 layers. Currently, the company has completed mass production of 218-layer products and Micron has completed mass production of 280-layer products. In his opinion, continuing to increase the number of layers is not a good strategy. As the product multiples increase, a large amount of funds will be required specifically for new equipment. However, if the market cannot digest the new products and starts to cut prices, it will form a vicious circle.

Li Yan then introduced some of the trade-offs of 3D NAND scaling. There are four main vectors of scaling, including vertical scaling, horizontal scaling, architectural scaling, and logical scaling. Taking the PC SN5000S NVMe SSD product as an example, she said that this product is a cost-effective, DRAM-free QLC PCIe Gen4 SSD using BiCS6 NAND technology. The internal controller and firmware provide a fully optimized solution that can be put on the market faster and quality controlled, providing 51% memory density and 12% layer density improvement compared to the previous generation; and compared with the mainstream TLC SSD, the performance of the QLC SSD has been improved, matching the TLC SSD in many indicators. In the consumer and even commercial fields, PC OEMs have adopted QLC SSDs in large quantities.

Gou Jiazhang, CEO of Silicon Motion Technology, said that AI will continue to drive global market growth. Storage is one of the key areas in the artificial intelligence ecosystem. The market needs more related software and applications to make AI edge devices more meaningful and attractive to consumers.

Ni Jinfeng, Vice President of Sales for Solidigm Asia Pacific: The company currently has leading innovative technologies in areas such as TLC and QLC, and can provide the industry with a powerful data center storage product portfolio, a storage product portfolio that optimizes AI storage efficiency, etc.

Liu Zhinong, Executive Vice President of Unigroup Spreadtrum: In the full-scenario AI computing system, software is the "engine", chips are the "base", ecology is the "link", and products are the "carrier". The company will continue to consolidate the advanced semiconductor intelligent manufacturing platform based on strategy and capability centers, maintain supply chain security, ensure efficient and stable product delivery, and improve customer satisfaction.

Xie Haibing, technical director of Intel's China Application Design Department, said that AI PC is the best carrier for edge AI and will lead the next explosion in the PC industry. Intel Core Ultra has powerful AI capabilities to realize the deployment of "generative AI" on PCs. The company has shipped 20 million AI PC chips so far, and it is expected to reach 40 million by the end of the year, and 100 million by 2024-2025.

Huang Haiqing, deputy general manager of R&D at Shenghong Technology, said that global technology companies are investing in AI at an unprecedented rate, and the development of AI applications is bound to drive the three major application markets of smartphones, PCs, and servers to continue to improve. Currently, memory bandwidth is growing, and the company is also accelerating its investment in product R&D to adapt to current application scenarios such as high-performance computing, data centers, AI training and reasoning. Currently, the fifth generation of high-performance memory (DDR5) is in mass production, and the sixth generation (DDR6) and seventh generation (DDR7) products are being developed.

Sun Chengsi, Chairman of BIWIN Storage, said that with the continuous development of the storage industry, the importance of packaging and testing has been increasing, especially as advanced packaging is developing towards miniaturization and integration, and the technical barriers are gradually increasing. The company is deepening its R&D packaging and testing integration layout to enhance its competitiveness. It is expected that the R&D packaging and testing integration 2.0 strategy will lead the company to upgrade from a storage product supplier to a full-service partner covering wafer-level advanced packaging and testing services, providing industry partners with higher-quality in-depth solutions and promoting the overall improvement of customer value.

Ding Rui, general manager of the product department of iFlytek's consumer platform business group, said that with the development of large models, AI has gradually changed from a "tool" to an "assistant", and can do more generalized things through a more natural dialogue. In the future, we hope to cooperate with all parties in the industry to promote the development of AI technology.

Ma Jian, president of business development at Arm's IoT division, said that the AI ​​wave is sweeping in, and generative AI is not only applied in the cloud, but also at the edge at an amazing speed, and storage plays a key role in AI computing from the cloud to the edge. For edge AI innovation in the new era, Arm is committed to advancing simultaneously in hardware, software, and ecosystem.