Why do SSDs wear out?

Every write forces electrons through a thin insulating layer, and each crossing damages that insulation slightly until the cell can no longer reliably hold a charge.

6 min read

Intuition
1

Simple intuition

The plain reason, in everyday words

An SSD stores each bit as a small amount of electric charge held in a tiny trap, surrounded by an insulating layer that keeps the charge from leaking away. To write new data, the drive has to force electrons through that insulating layer and into the trap — and to erase, it forces them back out again. Every time electrons are pushed through, they knock the insulating material about slightly, like footsteps wearing a groove in a stone step. One crossing does essentially nothing. A few thousand crossings leave the insulation leaky enough that the trap no longer holds its charge reliably, and the drive stops trusting that cell. Nothing is spinning, nothing is rubbing, and yet the drive genuinely wears out — because writing is a physically destructive act at the scale where the data lives.

What people get wrong

SSDs wear out from being read.

Reading is essentially non-destructive. Only programming and erasing push charge through the insulator. Read disturb exists but is a second-order effect handled by the controller.

Defragmenting an SSD helps performance.

There is no seek time to optimise, and defragmentation is a large volume of pointless writes that consumes real endurance. Modern operating systems disable it on SSDs deliberately.

When an SSD wears out it fails suddenly and without warning.

Cells degrade gradually and the controller reports remaining life through SMART attributes long before failure. Most worn drives switch to read-only rather than losing data outright.

A drive left in a drawer keeps data indefinitely because there are no moving parts.

Stored charge leaks over time, and it leaks faster in a heavily used drive and at higher temperatures. Flash is not an archival medium.

Why it matters

It changes how you use storage. It explains why filling an SSD to the brim slows it down and shortens its life, why swap-heavy configurations and constant logging are worth thinking about, why databases are tuned to write sequentially, and why the honest answer to 'is an SSD more reliable than a hard drive' is 'it fails differently'.

Where this came from

Who worked it out

Fujio Masuoka invented flash memory at Toshiba around 1980, and the NAND variant followed in 1987 — designed for density rather than durability, on the assumption it would replace film and tape rather than disks.

What problem forced it

The problem it solved was non-volatile storage with no moving parts, initially for cameras and portable devices where write volumes were tiny and endurance simply did not matter.

How it changed since

As flash moved into general-purpose computing, endurance became the binding constraint, and the response has been almost entirely at the controller level: wear levelling, TRIM, over-provisioning, and progressively more powerful error correction. Meanwhile the cells themselves got worse per-cell as the industry traded endurance for density, and 3D stacking in the mid-2010s was the first change to push back the other way.

Where to go next

Why batteries degrade

A different technology with the same shape of problem — normal use is physically destructive.

Why CPUs have cache

The layer above storage in the same memory hierarchy, with the opposite constraint.

Where this question came from

Written for Curio rather than collected from a forum — it is part of the curated corpus that ships with the platform. The references it draws on are listed under Sources.

Where curiosity goes next
See the map