SK Hynix Pushes 375 Layer NAND As The 400 Layer Race Hits A Material Wall

SK Hynix is moving closer to mass production of 375 layer 3D NAND by the end of 2026. According to The Elec, the company has completed production verification and is preparing to convert existing lines at its M15 fab in Cheongju instead of building a new facility. The 375 layer product was reportedly once described internally as 400 layer class NAND, but SK Hynix adjusted the final count due to manufacturing difficulty at extreme stacking levels. The company’s roadmap is still moving forward, with future 480 layer and 604 layer products expected to require a different process approach.

"Products once described as 400 layer class NAND were revised to 375 layers. The roadmap then extends to 480 layer and 604 layer products." Quote by: Industry source via The Elec

The bigger story is not only the layer count. It is the material change. SK Hynix is replacing part of the tungsten used in word lines with molybdenum, a move designed to reduce resistance as NAND structures become narrower and taller. As 3D NAND stacks increase, tungsten becomes harder to scale because resistance rises as wiring dimensions shrink. That can slow signal transmission and affect write and erase performance. Molybdenum offers lower resistance in fine word line structures and can also be deposited without the same auxiliary liner layer, helping improve density.

Samsung has already moved in this direction, using molybdenum in its ninth generation 286 layer V NAND and preparing its next generation 400 layer plus products. TrendForce also notes that the move toward 400 layer NAND is no longer only about stacking more layers, but about changing materials, bonding methods, and process architecture.

The material transition will also create a new supply chain pressure point. The Elec reports that Samsung used around 4 tons of molybdenum last year and has purchased around 10 tons this year, while SK Hynix is expected to use about 4 tons annually once its adoption ramps. Industry demand could rise to 25 tons in 2027, 40 tons in 2028, 60 tons in 2029, and 80 tons by 2030.

The NAND race is entering a more difficult phase. Adding more layers is no longer enough. Memory makers now need better materials, better etching, better deposition, and stronger process control to keep scaling alive. SK Hynix cutting its 400 layer class product down to 375 layers is not a failure. It is a sign that the industry is reaching a real engineering wall, where every extra layer increases risk, cost, and complexity.

Molybdenum may become one of the key materials for the next stage of NAND scaling. If SK Hynix and Samsung can make it work at high volume, higher density SSDs will keep moving forward. If not, the road to 500 layer and 600 layer NAND could become much slower and more expensive than expected.

For consumers, this will eventually matter in SSD capacity, efficiency, and price. For data centers, it could become even more important as AI storage demand continues to rise.

Do you think the next big SSD improvement will come from more NAND layers, new materials, or better controller technology?