There is a myth that has circulated since the early days of the SSD: that writing a lot “wears out” the drive and it will die within a couple of years. The truth is far more interesting. A modern SSD can survive hundreds of terabytes of writes (sometimes more than a petabyte) on memory that physically degrades with every erase. How? Through a hardware system that spreads the “fatigue” around, called wear leveling.
A NAND cell is a box of electrons
To understand it you have to go down to the physics. NAND flash memory stores data in cells, each able to hold one or more electric charges. A cell is essentially a transistor with a floating gate, a conductor insulated by a thin layer of oxide. When writing, the controller applies a voltage that pushes electrons through that oxide and traps them in the floating gate: that is a 1 or a 0 that persists without power.
Depending on how many bits each cell holds, we talk about SLC (1 bit), MLC (2), TLC (3) or QLC (4). More bits per cell makes the drive cheaper, but requires distinguishing increasingly finer charge levels, which reduces tolerance to wear and shortens lifespan.
Erasing is what wears things out
The wear mechanism works like this: each P/E cycle (program/erase), writing and erasing a cell, slowly breaks the oxide insulator. Every cycle erodes the barrier that holds the electrons. When the oxide degrades too much, the cell can no longer hold its charge and data becomes corrupted. By order of magnitude, an SLC cell typically survives tens of thousands of P/E cycles, while a QLC cell can fall below a thousand.
The asymmetry between writing and erasing
The key design point is that the operations are not symmetric. In an SSD you write by pages (about 4-16 KB) but can only erase by blocks (several megabytes), and a page that has already been written cannot be rewritten in place: the whole block must be erased first. To hide this limitation there is the FTL (flash translation layer), a table that maps the logical addresses seen by the operating system to the real physical positions in the memory.
When the system asks to overwrite a piece of data, the controller does not erase the old page: it writes the new one elsewhere and updates the map (a technique called copy-on-write). The old pages become “garbage” that a garbage collection (GC) process will clean up later.
Sharing the fatigue
This approach leads to the real problem. If the controller kept reusing the same block for new writes, that block would hit its P/E limit within days while the rest of the memory stayed pristine. Wear leveling prevents exactly that: it spreads writes across the whole chip so that every cell ages at the same rate.
There are two variants. Dynamic leveling acts during writes and moves hot data across still-fresh blocks. But data that never changes, such as an operating system, photos or backups, would sit forever on the same block, “freezing” its wear at zero. To compensate, static leveling goes further: it proactively moves that cold data into already-used blocks, and injects fresh blocks into the write circuit. The controller keeps per-block wear counters and tries to equalize their values.
The silent protagonist: TRIM
There is an actor that is not inside the chip but in the operating system. When you delete a file, the SSD has no reason to know those blocks are free: for it they still hold data. The TRIM command tells the controller that a range of blocks is no longer needed, so its garbage collector can free them without moving dead data. Without TRIM, the GC ends up copying garbage only to discover it was garbage, wasting wear and slowing the drive.
The picture is completed by over-provisioning: reserving a percentage of capacity (often 7-28%) that never shows up in the file system. That extra space gives the controller a cushion to keep the GC efficient and keep leveling wear without interrupting the user’s writes.
How an SSD’s lifespan is measured
Manufacturers publish the TBW (terabytes written): the total amount of data the drive can write before the warranty no longer covers it, and a good estimate of its useful life. A 1 TB drive with a TBW of 600 writes 600 TB; if you write 50 GB a day, there is theoretically more than 30 years left in it. Most users die before their SSD does.
In the end the paradox resolves itself: SSDs do not die from “writing so much” but from the impossibility of all cells degrading at once. Wear leveling is not magic: it is a constant orchestration of address mapping, garbage collection, TRIM and wear counters working together so that your drive ages in the most civilized way possible.
