3D NAND

3D NAND stacks NAND flash memory cells vertically (like a 3D apartment building) instead of horizontally. This dramatically increases density per mm². Leading-edge nodes in 2026: Samsung 286-layer V9 NAND, SK Hynix 321-layer, Micron 232-layer G9. Used in all modern SSDs · critical for AI training data storage where tens of petabytes per training run are common.

3D NAND density doubles every 2-3 years via layer-count increases and smaller cell sizes. PCIe 5.0 + NVMe 2.0 SSDs based on 3D NAND can read at 12-14 GB/s · sufficient for training data ingestion. Enterprise flash arrays (Pure Storage, NetApp, VAST) use 3D NAND to build petabyte-scale low-latency storage that feeds GPU training jobs. 3D NAND is distinct from HBM (AI's fast memory) · NAND is bulk storage, HBM is on-package working memory.

2026 node leadership: Samsung (V9 at 286L), SK Hynix (4b stacked at 321L), Micron (G9 at 232L), Kioxia-WD (BiCS 8 at 218L). CMOS-under-Array (CUA) architecture pushes density further. Interface: Toggle DDR 5.0 and ONFi 5.1 at 3.6 Gbps. Enterprise pricing: $60-100/TB for high-endurance NAND · dropping 20-30% per year. AI training pipelines now routinely use 100+ PB for frontier model training.