NAND Flash

Unplug your laptop and your SRAM cache and DRAM main memory both go blank instantly — but the files on your SSD are still sitting there the next time you boot up. That’s because SSDs (and USB drives, and the storage inside your phone) are built from NAND flash, a memory technology that stores each bit as trapped electric charge sealed inside an otherwise ordinary-looking transistor, with no power required to hold it in place.

The floating-gate transistor

A flash memory cell starts life as a regular MOSFET, then adds one crucial extra layer: a second gate, called the floating gate, sandwiched between the normal control gate and the channel, and completely surrounded by an insulating oxide layer.

Because the floating gate is fully surrounded by insulator, any electrons pushed onto it are physically trapped — with no conductive path in or out, they simply stay put, even with the power completely removed. That trapped charge is the whole mechanism: extra electrons sitting on the floating gate raise the voltage needed on the control gate to turn the transistor ON, so a cell can be read by checking whether it switches on at a particular test voltage. A cell with trapped charge (harder to turn on) reads as one logical value; a cell without it reads as the other.

🧩 Think of it like… a sealed glass jar of marbles buried in a wall, with a valve outside that only lets marbles in or out if you apply a lot of force (a high voltage) — everyday handling can't budge it. Whether the jar is full or empty doesn't depend on power being supplied; the marbles just sit there, sealed away, for years. Checking whether the jar is full or empty is a completely separate, gentle action from actually filling or emptying it.

Programming and erasing

Getting electrons onto or off of the floating gate requires forcing them through the thin tunnel oxide using a relatively high voltage — a physical process called Fowler-Nordheim tunneling (electrons are pushed through a barrier they normally couldn’t cross). Applying a strong voltage between the control gate and the channel pushes electrons onto the floating gate (programming, writing a 0 in the typical convention); applying the opposite strong voltage pulls them back off (erasing, writing a 1).

This tunneling process is exactly why flash memory can’t simply overwrite a single bit the way DRAM or a hard drive can. Erasing requires a relatively large, precisely controlled voltage pulse, and applying it to a single transistor without disturbing its densely packed neighbors isn’t practical — so NAND flash is only erasable in large groups called blocks (commonly hundreds of kilobytes to a few megabytes at once), even though it can be programmed (writing 0s) at a smaller granularity called a page (a few kilobytes).

This is why SSDs need a background process called garbage collection and a technique called wear leveling: to write new data to a partially-used block, the drive typically has to copy the still-valid data elsewhere, erase the whole block, and spread future writes evenly across all blocks — because each block can only endure a limited number of program/erase cycles before it wears out.

From cell to SSD

Individual flash cells are wired into long series chains — the origin of the “NAND” name, since reading through a chain electrically resembles the series-transistor structure of a CMOS NAND gate, even though the actual storage mechanism has nothing to do with logic gates. These chains are organized into pages and blocks, and modern chips often store more than one bit per cell by distinguishing multiple intermediate charge levels rather than just “charged” or “not charged” — 2 bits per cell (MLC), 3 (TLC), or 4 (QLC), trading some speed and durability for even greater density. An SSD controller sits on top of this raw array, translating the computer’s read/write requests into the flash-specific dance of programming, erasing, wear leveling, and error correction that the memory itself can’t handle alone.

Key takeaways

NAND flash stores a bit as electric charge trapped on an insulated floating gate — the charge stays put with no power applied, which is what makes flash non-volatile.
Getting charge onto or off the floating gate requires tunneling through a thin oxide layer using a relatively high voltage.
Because of how erasing works, flash can only be erased a whole block at a time, even though it can be programmed a smaller page at a time — the root cause of SSD garbage collection and wear leveling.
Compare this to SRAM and DRAM, both of which lose their data the instant power is removed.

DRAM Emerging Memory