SK Hynix begins mass production of the world's highest-end 321-layer NAND, completing its ultra-high-performance NAND memory portfolio.

SK Hynix announced on the 21st that it has started mass producing the world's highest-density 321-layer 1Tb (terabit) TLC (Triple Level Cell) 4D NAND flash.

SK Hynix stated, “We have been mass producing and supplying to the market the 238-layer NAND, the highest-layer product of the previous generation, in June 2023, and this time, we have overcome technological limitations by being the first to introduce a NAND that exceeds 300 layers,” and “We will respond to market demands by supplying 321-layer products to customers starting in the first half of next year.”

SK Hynix overcame the limitations of stacking by introducing the highly efficient ‘3-Plug’ process technology during the development of this product.

This technology divides the plug process into three stages, and then electrically connects the three plugs through an optimized subsequent process.

In the process, low-strain materials were developed and automatic alignment correction technology between plugs was introduced.

In addition, the company's technical team applied the development platform of the previous generation 238-layer NAND to 321-layer NAND, thereby minimizing process changes and improving productivity by 59% compared to the previous generation.

This 321-stage product has a data transmission speed that is faster than the previous generation.The read performance has improved by 12%, and the read performance has improved by 13%. In addition, the data read power efficiency has also increased by more than 10%. SK Hynix plans to actively respond to the new low-power, high-performance market for AI with 321-layer NAND and gradually expand the scope of its use.

SK Hynix Vice President Jeong-dal Choi (in charge of NAND Development) said, “As we have been the first to enter mass production of 300-layer or more NAND, we have gained an advantageous position to attack the AI storage market, including SSDs for AI data centers and on-device AI,” adding, “Through this, we will leap forward as a ‘full stack AI memory provider’ with a complete ultra-high-performance memory portfolio in both DRAM, represented by HBM, and NAND.”