Nothing is missing
Plug in a drive labelled "1 TB" and most operating systems report roughly 931 GB. No capacity has actually vanished. The drive contains exactly what the label says, and the "missing" 69 GB is entirely an artefact of two different definitions of the prefixes kilo-, mega-, giga- and tera- being applied to the same physical byte count.
Two prefix systems, same names, different meanings
| Decimal (SI-style) | Value | Binary (IEC) | Value |
|---|---|---|---|
| kilobyte (kB) | 1,000 bytes (103) | kibibyte (KiB) | 1,024 bytes (210) |
| megabyte (MB) | 1,000,000 bytes (106) | mebibyte (MiB) | 1,048,576 bytes (220) |
| gigabyte (GB) | 1,000,000,000 bytes (109) | gibibyte (GiB) | 1,073,741,824 bytes (230) |
| terabyte (TB) | 1,000,000,000,000 bytes (1012) | tebibyte (TiB) | 1,099,511,627,776 bytes (240) |
Storage manufacturers use the decimal column, and have done since long before the distinction had separate names, because it makes marketed capacity a round number and matches how every other industry uses "kilo" and "mega." A drive labelled 1 TB genuinely contains 1,000,000,000,000 bytes.
The exact arithmetic
Operating systems that compute "GB" using binary division, which is still the default behaviour of Windows File Explorer, take the drive's real byte count and divide by powers of 1,024, while continuing to print the label "GB" instead of the correct "GiB":
1×1012 bytes ÷ 230 bytes/GiB = 1,000,000,000,000 ÷ 1,073,741,824 = 931.32… GiB
That number, 931.32, is what a binary-dividing OS shows, mislabelled as "GB." Nothing was lost in formatting or overhead. It is the same 1,000,000,000,000 bytes, divided by a bigger divisor than the manufacturer used, and given the wrong unit name for the result.
Where the two-name system came from
The ambiguity is old. "Kilobyte" was used loosely for both 1,000 and 1,024 bytes from the earliest days of computing, because 210 (1,024) is close enough to 103 (1,000) that the sloppiness rarely mattered at small scales. It stopped being harmless once capacities reached the gigabyte range, where the two definitions diverge by 7.4%, and the terabyte range, where they diverge by almost 10%. In December 1998 the International Electrotechnical Commission approved Amendment 2 to IEC 60027-2, introducing the binary prefixes kibi-, mebi-, gibi- and tebi- specifically to end the ambiguity by giving binary quantities their own unambiguous names, leaving kilo-, mega- and giga- to mean strictly powers of 1,000, as they do everywhere else in metrology. That standard was later folded into IEC 80000-13 (2008), the part of the International System of Quantities covering information science and technology, which remains the current authority for both sets of prefixes.
Adoption was optional, and it shows
Standards bodies can define correct names; they cannot force operating systems to use them. Apple switched macOS to decimal-based reporting with Mac OS X 10.6 "Snow Leopard" in 2009 — a 1 TB drive shows as "1 TB" in Finder, matching the manufacturer's label, because Finder now divides by powers of 1,000. Windows never made the equivalent switch: File Explorer still divides by powers of 1,024 while printing the label "GB," which is why the same physical drive shows a different number depending on which OS it is plugged into. Linux is inconsistent by tool rather than by design — many command-line utilities, ls -lh and du -h without the --si flag among them, default to binary division, while several desktop file managers default to decimal, correctly labelled.
RAM is the one place binary is not just a labelling choice
Storage capacity is a manufacturing decision; a drive controller could in principle be built to expose any byte count at all. RAM capacity is not: memory chips are physically addressed with binary address lines, so a chip's capacity is architecturally a power of two. An "8 GB" RAM module genuinely contains 8 × 1,073,741,824 = 8,589,934,592 bytes, 8 GiB by the correct name, while an "8 GB" flash drive contains 8 × 1,000,000,000 = 8,000,000,000 bytes. Same printed label, two different physical quantities, for two different underlying reasons: one is a hardware constraint, the other is a units convention that a manufacturer chose.
Bits, not bytes, for network speed
One more prefix trap sits next to this one: network speeds are quoted in bits per second, file sizes in bytes. Since a byte is 8 bits, an advertised 1 Gbps connection tops out around 125 MB/s of file transfer before protocol overhead, not 1,000 MB/s and not 1 GB/s. Dividing an advertised link speed by 8 to estimate real file-transfer throughput resolves most of the confusion between "the connection is fast" and "the download feels slow."
The one format that used neither definition consistently
The 3.5-inch "1.44 MB" floppy disk is the odd one out, and it is a genuinely instructive case rather than a footnote. Its actual formatted capacity is 1,474,560 bytes. That number is not 1.44 × 1,000,000 (1,440,000, the decimal reading) and it is not 1.44 × 1,048,576 (1,509,949, the binary reading). It is 1.44 × 1,024 × 1,000: a hybrid where "kilo" is read as the binary 1,024, matching how memory and disk sectors are actually addressed, but "mega" is then read as a decimal 1,000 multiple of that already-binary kilobyte. Neither the IEC binary convention nor the plain SI decimal convention describes it; the format predates both the 1998 standardisation effort and the marketing pressure that came with multi-gigabyte drives, and it shipped with a unit definition that satisfied nobody in particular.
A quick way to sanity-check any advertised size
If a number is being used to describe hardware capacity you can physically count in binary units, such as RAM or flash chip die size, expect it to be a clean power of two and therefore identical whichever prefix system is used. If it describes something manufactured or provisioned to a round decimal target, storage drives, cloud storage tiers, bandwidth allowances, expect the "missing" percentage to show up the moment an operating system reports it back to you in binary-divided, mislabelled units.