Northeast Securities Co.,Ltd. released a research report stating that the global data center storage demand market exhibits exponential expansion characteristics, forming a dual evolution trend of "quantum leap + accelerated growth." By 2024, data volume is expected to surpass 1.1ZB, projected to rise to 2.4ZB by 2028. As data becomes the foundation for AI training and inference, its significance has substantially increased. Storage has evolved into a critical variable determining AI efficiency, with storage technology directly impacting data scale, access speed, cost, and AI model training effectiveness. Storage is no longer merely a data repository but a decisive factor in AI performance, driving storage systems into a new phase of multi-dimensional optimization balancing "capacity, performance, and efficiency." Key insights from Northeast Securities include:
1. AI Era Transition: Large-Scale "Cold Data" Turns "Warm," Creating New Opportunities for Storage Media With growing demand for data access in large-model training and inference, previously classified "cold data" is being reactivated. These datasets, now frequently used in model iterations and real-time inference, are transitioning into "warm data" or even "hot data" due to continuous access. This shift significantly improves data utilization efficiency while imposing new requirements on storage media. Consequently, as cold-to-warm data conversion accelerates, SSD and innovative media like MED are expected to see expanded applications.
2. Data Explosion Scenario: Intelligent Tiered Architecture Emerges as Critical Evolutionary Direction The exponential growth in data volume presents multiple challenges for lifecycle management: - Escalating storage costs for massive datasets - Compromised data access efficiency - Difficulties ensuring data security and compliance
Intelligent tiered architecture, with its scalability, flexibility, and multi-data-type adaptability, has become the next-generation storage foundation. The cold-hot data tiering strategy categorizes data by access frequency, storing them on media with corresponding performance and cost profiles: - Hot data (frequent access): High-performance, high-cost media - Cold data (infrequent access): Low-performance, low-cost media
Advantages include: - Reduced storage costs - Enhanced access efficiency - Optimized resource allocation - Improved data security
3. Magnetoelectric Storage: Hybrid Architecture Outperforming Traditional HDDs, Potential Optimal Solution for Warm-Cold Data MED technology focuses on upgrading data center storage systems, particularly replacing traditional HDDs in warm-cold data scenarios. Through innovative hybrid design, MED integrates SSD-level speed with tape storage's massive capacity within a single sealed device, delivering comprehensive improvements in performance metrics, storage capacity, and cost control.
Risk factors: Industry cyclical downturn, competitive pressures, upstream supply chain risks, technological advancement delays, downstream client concentration risks, and potential shortfalls in earnings forecasts and valuation assessments.