Guosen Securities released a research report highlighting key technology trends in the storage market, including HDD, SSD, NAND, DRAM, and HBM, with numerous major players involved. The rapid growth of AI model inference is driving substantial demand for storage. According to the firm's estimates, AI inference will require 23.0EB of DRAM and 593.5EB of NAND by 2026, creating a short-term supply shortage that is expected to push storage prices higher, signaling the ongoing strength of the storage supercycle.
Key insights from Guosen Securities include:
**Storage System Composition and Classification** 1) **System Structure**: AI storage systems are divided into network-side storage (primarily for cold data, using HDD and SSD) and local storage (for hot and warm data, utilizing HBM, DRAM, and local SSD). 2) **Storage Categories**: Storage can be classified into volatile memory (Memory, including DRAM and HBM, known for speed) and non-volatile storage (Storage, such as SSD and HDD, offering high capacity and low cost).
**Market and Technology Trends** 1) **HDD**: Hard disk drives improve capacity through increased areal density, with HAMR (Heat-Assisted Magnetic Recording) technology significantly boosting single-disk capacity. The market is dominated by Seagate and Western Digital. 2) **SSD**: Solid-state drives, based on NAND Flash, continue to advance in stacking layers and transitioning from SLC to MLC and QLC architectures. Major players include Samsung, SK Hynix (including Solidigm), Micron, SanDisk, and Kioxia. 3) **DRAM**: With advantages in low power consumption, high integration, and cost efficiency, DRAM has evolved from SDRAM to DDR, with leading manufacturers now developing DDR6. Key applications include smartphones, PCs, and servers, with Samsung, SK Hynix, and Micron as primary suppliers. 4) **HBM**: High Bandwidth Memory stacks multiple DRAM chips vertically via TSV interconnects, delivering higher density and bandwidth, primarily used in AI. Major suppliers include SK Hynix, Micron, and Samsung. 5) **NAND**: Increasing bit density and layer stacking support SSD and smartphone demand, with Samsung, SK Hynix, Kioxia, Micron, and SanDisk leading the market.
**Demand Projections** AI model inference is accelerating storage demand. By 2026, AI inference is projected to require 23.0EB of DRAM and 593.5EB of NAND. The current supply-demand imbalance is expected to sustain rising storage prices, reinforcing the storage supercycle.
**Leading Storage Companies** - Samsung Electronics and SK Hynix dominate DRAM, HBM, NAND, and SSD segments. - Micron offers a comprehensive product portfolio but trails slightly in market share. - SanDisk and Kioxia specialize in NAND and SSDs, while Western Digital and Seagate focus on HDDs.
**Risk Factors** - Potential price declines due to DRAM/NAND capacity expansion. - Lower-than-expected capital expenditures from major internet firms. - Slower-than-anticipated growth in AI application users. - AI model optimization reducing storage demand.
Comments