SSDs stack NAND flash in dozens to hundreds of vertical layers — 3D NAND — because engineers ran out of room to keep shrinking flash cells sideways without cells interfering with each other. Building upward instead let manufacturers keep adding capacity affordably, which is why SSD prices per gigabyte have fallen for a decade.
The Problem: Planar (2D) NAND Hit a Wall
For its first two decades, NAND flash was built as a flat grid of cells on a single layer of silicon, and each new generation simply shrank the cells to fit more into the same chip area — the same scaling strategy used across the whole chip industry, covered in general in our how semiconductors are made explainer. But flash cells behave differently from logic transistors: as planar cells shrank past roughly the 15-nanometer mark, they packed so close together that the electric field of one cell started to bleed into its neighbors, corrupting nearby charge levels — a problem called cell-to-cell interference. At the same time, each cell's floating gate held fewer and fewer electrons, making the difference between a correctly and incorrectly read bit razor-thin. Planar NAND had reached a genuine physical dead end for further shrinking.
The Fix: Build Vertically Instead of Shrinking Further
3D NAND sidesteps the problem by turning the layout on its side. Instead of shrinking cells to pack more into a flat plane, manufacturers stack dozens to hundreds of layers of cells vertically on the same chip footprint, using cells that are actually larger and less interference-prone than the smallest planar designs ever managed. Capacity comes from height, not from cramming cells closer together — which relaxed the interference and reliability problems that planar scaling had run into, while still growing storage density generation after generation.
What a "Layer" Actually Is: Vertical Strings of Cells
Picture a tall, narrow cylindrical channel of silicon drilled straight down through dozens of stacked, alternating layers of conductive and insulating material. Where each conductive layer wraps around that channel, it forms one flash cell — the layer acts as the cell's control gate, and a thin charge-trapping layer between the channel and the gate stores the charge (a close cousin of the floating-gate mechanism described in our flash memory explainer, adapted for this vertical geometry). One drilled channel running through, say, 176 layers forms a "string" of 176 cells stacked on top of each other. Multiply that string across millions of positions on the chip, and you have a complete 3D NAND die.
Why More Layers Means More Capacity Per Chip
Each additional layer in the stack adds one more cell to every vertical string on the die, at essentially the same silicon footprint — which is why manufacturers have raced from roughly 32 layers in early 3D NAND to well over 200 layers in current generations. That layer count, multiplied by how many bits each individual cell holds (see our SLC vs. MLC vs. TLC vs. QLC breakdown), and multiplied again by how many dies a manufacturer stacks inside one physical chip package, is what determines a modern SSD's total capacity. It's why terabyte-class NVMe drives now fit comfortably on a stick the size of a stick of gum.
The Tradeoffs of Going Tall
Stacking layers is not free. Drilling a perfectly straight, uniform channel through 200-plus layers is a serious manufacturing feat — any tilt or variation in the hole's diameter changes the electrical characteristics of cells at different heights in the same string, which is why controllers apply per-layer calibration to compensate. More layers also means more complex, more expensive fabrication steps, even though the payoff in density is worth it. And because all the cells in a string share the same vertical channel, certain operations touch the whole string at once, which shapes how controllers organize pages and blocks internally.
Where 3D NAND Shows Up
Virtually every SSD sold today, from budget SATA drives to the fastest NVMe sticks, is built on 3D NAND — planar NAND is essentially obsolete for anything but legacy or ultra-low-capacity parts. High-capacity, high-speed NVMe drives make the layer count's payoff most visible: a Crucial P310 1 TB Gen4 NVMe SSD and a WD_BLACK SN7100 1 TB NVMe SSD both pack a full terabyte onto a single M.2 stick by combining many stacked layers with multiple bits per cell, while a workstation-grade drive like Samsung's 970 PRO shows the other side of the tradeoff — slightly lower per-cell density in exchange for the speed and endurance headroom that comes from a more conservative bits-per-cell choice on the same layered foundation.
Frequently Asked Questions
Is 3D NAND better than planar NAND?
For essentially every practical purpose, yes. It scales to far higher capacities without hitting the cell-to-cell interference and reliability problems that stopped planar NAND from shrinking further, which is why it has replaced planar NAND across almost the entire SSD market.
How many layers do modern SSDs have?
It varies by manufacturer and generation and changes frequently as new products ship, but the industry has moved from roughly 32 layers in early 3D NAND to well over 200 layers in current-generation flash, with the count continuing to climb.
Does more layers mean a faster SSD?
Not directly — layer count mainly drives capacity per chip. Speed depends more on the controller, the interface (SATA vs. NVMe, and which PCIe generation), and how many bits are packed into each cell.
Why don't manufacturers just keep shrinking planar NAND instead?
Because shrinking flat cells past a certain point causes neighboring cells to electrically interfere with each other and leaves too little charge margin to reliably distinguish a 0 from a 1. Building vertically avoided that wall entirely rather than trying to push past it.
Share this article: Twitter