The drive maker counts a terabyte in decimal terms and many operating systems display in units of 1024 while keeping a decimal-looking suffix. Two conventions are being used for one drive, and nothing has been lost.
Two different things reduce the number on the screen, and only one of them is a reduction at all.
The unit convention, which takes nothing away
Storage makers count in decimal terms: a kilobyte is a thousand bytes, a megabyte a million, a gigabyte a thousand million, a terabyte a million million. That is the ordinary meaning of the prefixes, and it is how capacity is printed on the box and on the label.
Many operating systems count in units of 1024 instead, because that is the natural granularity of addressing. Some of them label the result with the IEC binary suffixes; a great many keep the decimal-looking suffix while doing binary arithmetic, and that is the whole source of the confusion.
The consequence is that a drive displays as a smaller figure than the box, and the discrepancy grows with capacity, because each successive prefix multiplies the difference again. This site reports the figure the maker publishes, in the unit the maker publishes it in, and does not restate it in the other convention — the two are not the same quantity and converting one into the other invents a number the maker never gave.
The two conventions are told apart by suffix where anyone bothers: TB against TiB, GB against GiB. Where a display uses the decimal suffix and the binary arithmetic, no suffix will tell you, and only the arithmetic will.
Formatting, which does take something away
This one is real. A filesystem writes structures of its own before it stores a single file: a record of where things are, allocation information, journals, and reserved space on some filesystems.
How much depends on the filesystem chosen and on the parameters it was created with. It is a genuine reduction in what you can fill, it is small relative to the whole on a large drive, and it is separate from the unit question that gets blamed for it.
The other reductions, where they exist
A recovery partition on a machine that came with one occupies part of the drive and is usually not visible as free space.
Overprovisioning on a solid-state drive is capacity the controller keeps for its own use — wear levelling, spare blocks, garbage collection. It is set aside before the drive is presented, so the advertised capacity is already net of it rather than reduced by it afterwards.
A redundancy level in a multi-drive array is by far the largest of these, and it is not a hidden cost at all: a mirrored pair presents the capacity of one member, and a parity array gives up the equivalent of one or two members depending on the level. That is the arrangement working as specified.
How to compare two drives honestly
Compare the maker’s figures against each other, in the maker’s unit. Two drives quoted the same way are comparable; a maker’s figure held up against a screenshot of an operating system is not, because the two are counting differently.
And when planning space, plan in whichever unit your machine displays, since that is the number you will be watching fill up. Just do not use it to decide that a drive was mis-sold.
The questions that come up before an order
Is the drive maker allowed to count it that way?
It is the ordinary meaning of the prefix. Everywhere outside computing, a kilo is a thousand and a mega is a million, and storage makers use the prefixes in that sense. The binary interpretation is the one that is unusual, and the IEC published separate prefixes so that the two could be told apart.
Does formatting take some capacity as well?
Yes, and it is a separate matter from the unit question. A filesystem writes its own structures before it holds any of your files, and how much they occupy depends on the filesystem and on how it was created. That is a real reduction; the unit difference is not.
Last reviewed 10 September 2026