With the theme of "AI Leads Memory's New Momentum", this year's summit has brought together leading enterprises from various segments of the memory industry chain, including major wafer manufacturers, memory controller suppliers, module producers, packaging and testing enterprises, equipment and materials manufacturers, and terminal manufacturers, as well as representatives from investment institutions. The event centered around all aspects of key topics such as product innovation, technological evolution, and the collaborative development of the industrial chain, all with the aim of advancing industry cooperation and creating a shared future of success.
Accelerating Deployment of AI Applications to Unlock New Growth in the Storage Industry
The explosive growth of AI in the new era is reshaping the global semiconductor landscape. The widespread application of AI is driving demand for high-performance, high-computing-power chips, while also creating significant growth opportunities for high-bandwidth, large-capacity, and low-power storage chips.
The GMIF2024 Innovation Summit officially kicked off with the welcoming remarks from Mr. Rixin Sun, President of the Shenzhen Memory Industry Association (SMIA). Over the recent years, the exponential growth of AI, 5G, Big Data and Internet has seen a brand-new revolutionizing period in global storage industry. Mr. Sun addressed that the GMIF2024 Innovation Summit turns special attention to the entire upstream and downstream storage value chain that goes beyond just the storage and memory itself. In particular, it unfolds critical areas across the industry, consisting of storage media, solutions, system platforms, testing equipment and more. Mr. Sun affirmed that only through collaboration across the entire industry chain can we continuously drive the advancement of the sector and maintain a leading position in global competition.
Xiaofei Zhang, Chief Analyst for Electronics Industry from Haitong Securities, stated that, from the perspective of overall storage market, the HBM3e has exerted impacts on the production schedule of DDR5, and the decline in DRAM prices remains restricted. In the NAND sector, the inventory adjustments at server terminals are entering the final stages; coupled with AI driving demand for high-capacity storage products, prices have continued to rise in Q2. Although the recent increase in NAND prices has slowed, it still exceeds market concerns. Looking ahead, the penetration of AI terminal applications is accelerating, leading to a sustained increase in demand for computing and storage solutions.
Yimao Cai, Dean of the School of Integrated Circuits at Peking University, expressed that semiconductor storage, considered as one of the largest segment of the integrated circuit industry, has encountered considerable challenges after stepping into the post-Moore and AI era. On one hand, traditional memory technologies have posed significant limits on the integration density and reliability below 28nm, and there is an urgent need for breakthroughs in underlying technologies. On the other hand, the traditional computing chips are insufficient to satisfy the high-efficiency demands for intelligent equipment, catalyzed by the surging growth in AI computing needs, as the traditional chip hardware incurs high costs and energy consumption. It can be observed that innovations in foundational units and process integration are crucial to the progress of memory technology, and new forms of storage technology are becoming increasingly mature.
Prasad Alluri, VP and GM for Client Storage, SBU from Micron Technology, emphasized that "AI is everywhere and AI applications have entered every corner of people's daily lives in all kinds of forms. For instance, smartphone manufacturers have introduced AI capabilities in high-end phones, and the storage capacity of LPDDR 5 has been increased to between 12GB and 16GB to accommodate ever-increasing datasets. In addition, storage technology has evolved to UFS 4.0, which more than doubles power efficiency compared to the previous generation. In the realm of edge devices, autonomous driving will be elevated from Level 0 to Level 5 supported by the enhanced AI technology. Industry experts predict that the bit density required for automotive memory will increase by approximately 30 times, while the non-collision bits in power systems will grow nearly 100 times. None of the enhancements, whether in data centers, edge devices, or memory and storage environments, would be possible without the technological progress we've achieved generation after generation."
Yan Li, Vice President of Advanced Technology at Western Digital, pointed out that the 3D NAND flash market is highly competitive and constantly evolving. The number of NAND cells stacked has experienced substantial growth, from 48 layers in 2017 to 64 and 96 layers. Up to date, Western Digital has completed mass production of its 218-layer products; while Micron Technology has already mass-produced its 280-layer products as stated in its keynote. Dr. Yan Li shared her insights that the continuous increase of layers might not be considered as an ideal strategy because large amounts of capitals are required for new equipment as complexity of the products rises. Should the market fail to accommodate these new products and prices start to decline, it could lead to a vicious cycle.
Dr. Yan Li then moved on to make some explanations of the trade-offs involved in 3D NAND scaling technology. There are four main vectors to scaling: vertical scaling, lateral scaling, architecture scaling, and logical scaling. Taking its product PC SN5000S NVMe SSD as an example, Dr. Li noted that this model is equipped with cost-effective BiCS6 NAND technology and DRAM-less QLC PCIe Gen4 SSD, and its internal controller and firmware provide a perfectly optimized solution which can be introduced to the market within a shorter time and allowed for quality control. Compared to the previous generation, it offers a 51% improvement in memory density and a 12% increase in layer density. In relation to the mainstream TLC SSD, QLC SSD features enhanced performance and its major metrics are matched with TLC SSD. In consumer and commercial sectors, QLC SSDs are now being widely adopted by PC OEMs.
Wallace C. Kou, President & CEO of Silicon Motion, claimed that AI will be the ever-lasting energy and vitality that propel continuous growth in the global market. Storage is regarded as one of the key areas in the artificial intelligence ecosystem, and the market requires more relevant software and applications to make AI edge devices more meaningful and appealing to consumers.
Benny Ni, GAR Sales VP at Solidigm, remarked that the company currently possesses leading innovative technologies in the fields of TLC and QLC, enabling it to offer a powerful portfolio of data center storage products and storage solutions that optimize AI storage efficiency.
Zhinong Liu, Executive Vice President of UNISOC, signified that in the all-scenario AI computing system, software serves as the "engine", chips as the "foundation", the ecosystem as the "link", and products as the "carrier". The company will continue to strengthen its advanced semiconductor manufacturing platform based on strategy and capability centers, maintain supply chain security, ensure efficient and stable product delivery, and enhance customer satisfaction.
Haibing Xie, Director of Application Design in Center, Intel China, mentioned that AI PCs are the best carriers for edge-side AI and will lead the next explosion in the PC industry. The Intel Core Ultra is equipped with powerful AI capabilities, enabling the deployment of "generative AI" on PCs. At present, the company's AI PC chips have shipped 20 million units, and it is expected to reach 40 million units by the end of the year, with projected shipments of 100 million units in 2024-2025.
Haiqing Huang, R&D Vice General Manager at Victory Giant Technology, revealed that the AI investment landscape among global tech companies is currently marked by an unparalleled enthusiasm, and the three main application markets, namely the smartphones, PCs, and servers, are bound to experience a welcome upswing brought by the development of AI applications. As of now, there is a constant increase in memory bandwidth, and Victory Giant has been accelerating its product R&D investments to meet the demands of high-performance computing, data centers, AI training, and inference scenarios. The fifth-generation high-performance memory (DDR5) is already in mass production, while the development of the sixth generation (DDR6) and seventh generation (DDR7) products is underway.
