The memory semiconductor industry is experiencing a boom driven by artificial intelligence (AI), benefiting not only DRAM but also NAND flash. Samsung Electronics and SK Hynix anticipate continued strong demand for NAND and are accelerating the development of next-generation products.

● Memory Industry, Next-Generation AI NAND Boom

SK Hynix recently announced plans to build a next-generation AI-NAND (AIN) lineup, including high-bandwidth flash (HBF), at the '2025 OCP (Open Compute Project) Global Summit' held in San Jose, California. HBF is a product made by stacking NAND, similar to high-bandwidth memory (HBM), which stacks multiple DRAMs.

Until now, NAND has developed by stacking as many storage cells as possible. Therefore, HBF, which stacks completed memory chips, represents a new approach. While HBM specializes in data processing speed, HBF focuses on large-capacity storage. SK Hynix explained, "As the AI market grows, the demand for NAND that can efficiently process large amounts of data is increasing."

SK Hynix has named this product 'AI NAND B,' taking the letter 'B' from Bandwidth. Additionally, they plan to launch three lineups: 'AI NAND D (Density),' characterized by low power, low cost, and large capacity, and 'AI NAND P (Performance),' which has strengths in data processing speed.

According to the semiconductor industry, Samsung Electronics has also begun developing HBF targeting the AI server NAND market, although it has not made an official announcement. Samsung Electronics is attempting to differentiate itself from competitors by incorporating foundry (semiconductor contract manufacturing) production methods into memory.

Samsung Electronics is expected to apply the FinFET process, an advanced foundry technology, to NAND. FinFET is a technology used in ultra-fine processes ranging from 3 to 14 nanometers (nm, 1 nm is one-billionth of a meter). When applied to NAND, it can increase semiconductor integration, thereby enhancing performance and power efficiency. The finer the process in semiconductors, the more information can be stored and processed in the same space.

● AI-Driven Benefits Spread from DRAM to NAND

Unlike DRAM, which benefited from the AI boom, NAND suffered from oversupply and inventory accumulation since 2022. While DRAM is specialized for data processing speed, NAND focuses on capacity, leading to differing fortunes for the two memory product lines. The AI industry has concentrated on quickly performing calculations with low power, leading to increased demand for DRAM.

However, as competition among big tech companies in AI intensifies, demand for high-value-added NAND products is also increasing. An industry official stated, "As the ability to efficiently manage large amounts of data in AI servers becomes increasingly important, the need for NAND upgrades is growing." Analysts suggest that while Samsung Electronics recorded an 'earnings shock' in the second quarter (April–June) due to NAND inventory, the better-than-expected performance in the third quarter (July–September) was influenced by increased profits from NAND sales.

Popular News