TrendForce's latest research, in the past two years, the focus of AI-based infrastructure construction has been more on supporting high-efficiency recommendation services. Traditional large-capacity HDDs are in a serious shortage. CSP operators...
TrendForce's latest research, in the past two years, the focus of AI-based infrastructure construction has been more on supporting high-efficiency recommendation services. Traditional large-capacity HDDs are in a serious shortage. CSP operators have been turning to NAND Flash suppliers to solvers, giving birth to Nearline SSDs (nearline solid-state hard drives) designed to recommend AI design to meet the urgent market demand.
To fill in the supply gap, major NAND Flash suppliers have accelerated the introduction of near-line QLC NAND Flash product verification. QLC can store more data at a lower cost, becoming a key to meeting large-capacity needs. In addition, suppliers have also expanded QLC SSD production capacity and will gradually increase capacity utilization in 2026. After the AI application is expanded, the demand surge will continue until 2027, so the supply of enterprise-level SSDs in 2026 is in a tight state.
TrendForce said that in order to expand the advantages of near-line SSDs in AI storage applications and replace HDD more effectively, products will develop towards larger capacity and lower costs. The manufacturer is actively researching and launching new near-line SSD products, not only exceeding mainstream HDDs in capacity, but also optimizing costs, while significantly reducing power consumption.
NAND Flash application diversification: training and recommendationsIn addition to recommending AI applications, NAND Flash suppliers have also simultaneously published HBF (High Bandwidth Flash) products to expand the market share of AI training applications, forming two technical routes. SanDisk-led construction, a hybrid design combining HBM and HBF, aims to combine large capacity and extreme performance, and meet the dual requirements of AI model training for data throughput and capacity.
The enterprises represented by Samsung and the enterprise are SCM (storage memory) XL-Flash and Z-NAND, etc. Trials offer more cost-effective solutions than HBM to attract a larger customer base.
TrendForce said that this technological competition has also promoted the integration of NAND Flash applications from single pure storage to deeper AI computing, indicating that NAND Flash applications will be more diverse in the future.
