SK Hynix begins mass production of 321-layer QLC NAND flash, making a full-scale attack on the AI data center market.

SK Hynix announced that it has completed the development of a 321-layer 2Tb QLC NAND product that applies QLC (Quadruple Level Cell) technology to NAND flash with 300 or more layers for the first time in the world, and has begun mass production on August 25.

The 321-layer QLC NAND stores 4 bits in a single cell, doubling the storage capacity compared to existing products, and is scheduled to be supplied to global customers' AI data center markets starting in the first half of next year.

This 321-layer QLC NAND has solved the speed reduction problem that typically occurs with capacity expansion by expanding the 'plane' structure.

SK Hynix maximized parallel processing by increasing the number of planes capable of independent operation within NAND from four to six.

This improves data transfer speeds by 100%, write performance by up to 56%, and read performance by 18%.

Power efficiency also increased by more than 23%, securing competitiveness in low-power, high-performance environments such as AI data centers.

SK Hynix plans to expand the scope of its application to include not only eSSD for data centers but also UFS for smartphones and SSD for PCs.

SpecialThe company is preparing an ultra-high-capacity eSSD lineup that doubles the integration compared to existing products by introducing a unique stacking method that stacks 32 NANDs in a single package using 32DP (32 Die Package) technology.

Through this, the company presented a blueprint to leap forward as a 'Full Stack AI Memory Provider' in the storage market for AI servers.

“In order to respond to the explosively increasing demand for AI and the high-performance, high-capacity needs of data centers, we have begun mass production of 321-layer QLC NAND that has both price competitiveness and technological prowess,” said Jeong Woo-pyo, Vice President of SK Hynix (NAND Development). “We will significantly strengthen our high-capacity product portfolio to lead the global AI memory market.”