The significance of high-added-value Quad Level Cell (QLC) NAND products in the artificial intelligence (AI) memory market is rapidly increasing as global big tech companies adopt high-capacity Solid State Drives (SSDs) in the process of building their own AI servers. This trend is fueling expectations of explosive demand growth similar to last year's Graphics Processing Unit (GPU) driven surge in AI semiconductors centered around High Bandwidth Memory (HBM).
According to market research firm Omdia on June 10, the QLC NAND market size is expected to grow by 85% this year compared to last year, with its market share in the overall NAND market increasing nearly 8 percentage points from 12.9% last year to 20.7% this year.
Omdia forecasts that by 2027, QLC NAND will account for 46.4% of the total NAND market, more than doubling its share in just three years, approaching that of Triple Level Cell (TLC) products, which currently dominate the market with a 51% share. Notably, whereas QLC NAND products were primarily consumer-oriented until last year, this year the demand is expected to grow mainly for higher-priced server-oriented products.
QLC technology allows for the storage of four bits per cell, compared to three bits for TLC, and two and one bit for Multi-Level Cell (MLC) and Single-Level Cell (SLC), respectively. This means QLC NAND can further increase storage capacity compared to other NAND types.
The characteristics of QLC NAND align well with the needs of big tech companies intending to deploy generative AI on their own servers. The adoption of SSDs, which offer faster data reading and writing speeds than traditional Hard Disk Drives (HDDs), highlights the advantages of more information per unit area and reduced power consumption.
The price gap between SSDs and HDDs, previously a barrier to widespread adoption, has also been closing rapidly. From a 20-fold difference in 2015, the price difference has narrowed to 10-fold this year, with the general downturn in the NAND market since the second half of 2022 leading to reduced SSD costs. The slower writing speed and shorter lifespan, traditionally seen as disadvantages of these products, have been offset by data centers shifting to read-intensive workloads.
NAND manufacturers are also responding quickly to the surge in demand for QLC NAND. There is optimistic talk that the 'spring of NAND' might be stronger than expected. Like the demand growth for HBM last year based on AI demand, the NAND market could experience a similar long-term market formation.
In the semiconductor industry, Solidigm, a NAND subsidiary of SK hynix, is expected to benefit significantly. As of the end of last year, 60% of Solidigm's total production volume was QLC NAND. While Micron (40.1%) has a higher market share than Solidigm (21.1%), Micron's share is mainly in consumer products.
The supply of high-added-value QLC SSDs for AI servers is likely to be led by Solidigm and Samsung Electronics. Samsung Electronics plans to mass-produce the 9th generation 280-layer NAND in QLC format in the second half of this year. SK hynix, through Solidigm, is introducing a 60-terabyte (TB) QLC product and is also preparing a 300TB product for next year to maximally meet the AI-oriented demand.
|