Cookies for analytics and advertising
We use cookies for analytics and advertising, both sent to Google. Refusing changes nothing you can see.Read the privacy page
1 kB = 1e-9 TB
A terabyte is a thousand million kilobytes, so converting kB to TB moves the decimal point nine places. Almost nothing is measured in kilobytes and also worth quoting in terabytes, which means this conversion is nearly always the end of a multiplication — a small per-item size against a very large count.
500 kB is 5e-7 TB
— a lean web page.
64 kB is 6.4e-8 TB
— a long email.
1000000000 kB is 1 TB
— a drive as the box describes it.
8000000000 kB is 8 TB
— a large desktop drive.
| kB | TB |
|---|---|
| 1000000 | 0.001 |
| 2000000 | 0.002 |
| 5000000 | 0.005 |
| 10000000 | 0.01 |
| 50000000 | 0.05 |
| 100000000 | 0.1 |
| 500000000 | 0.5 |
| 1000000000 | 1 |
Convert kB to TB
A kilobyte is 1,000 bytes when a standards body defines it and 1,024 when an operating system does. This page uses 1,000; the kibibyte is the unit that means 1,024.
A terabyte is a trillion bytes. A drive sold as 1 TB is exactly that — the space that seems to go missing is a unit disagreement, not a defect.
Going from kilobytes to terabytes moves the decimal point 9 places to the left and changes nothing else. There is no factor to remember and no rounding to decide: the digits stay in the same order and only their position changes.
1,234 kB is 0.000001234 TB — same digits, moved along. That is worth knowing because it is the one kind of conversion you can check at a glance: if the digits of the answer are not the digits you started with, something other than the conversion has happened to them.
A terabyte is 1,000,000,000 kilobytes, a factor of a thousand million. No other conversion in this category spans so far, and the distance is the reason the page exists: the two units belong to different conversations. Kilobytes describe one small thing that somebody could open. Terabytes describe an amount of storage that has to be bought, racked or rented, and nobody in that conversation cares about a kilobyte.
What bridges them is always a count. Ten million thumbnails at 100 kB, a billion log lines at a kilobyte, four hundred million messages at 2.5 kB — each of those is a terabyte, and each is a sentence that says more than the converted figure does. When the conversion is run without a count behind it, the answer is a decimal with eight leading zeros and it tells nobody anything.
The shape of the real problem is one small number and one large one, and the large one is where the risk sits. A per-item size can usually be measured directly: take a thousand records, look at the bytes, divide. A count is a forecast, and forecasts of counts move by factors rather than by percentages. Doubling the retention period or the number of tenants doubles the answer; getting the per-record size wrong by a fifth does not.
That asymmetry is worth carrying into how the estimate is written down. "Forty kilobytes a record, three hundred million records a year, two years retained" is auditable — somebody can disagree with one number without discarding the whole thing. "24 TB" is not, and it is the figure that ends up in a budget, where nobody can see which of the three inputs it was most sensitive to.
A kilobyte figure that was rounded, multiplied by a count that was estimated, produces a terabyte figure that looks exact because the arithmetic was exact. It is not. If the per-item size is good to a fifth and the count to a factor of two, the answer is good to a factor of two, and printing it as 4.3271 TB misrepresents that more thoroughly than any rounding error could.
The discipline is to carry the uncertainty rather than the decimals. Round the answer to one significant figure more than the least certain input, say what the inputs were, and give a range where the count is genuinely unknown. Four to nine terabytes is a useful thing to tell somebody who has to buy storage; 4.3271 TB is a number they will plan against and then be surprised by.
A measured record size is the payload and rarely the cost. Databases round rows up to page boundaries, keep indexes that on a heavily queried table can match or exceed the data itself, hold write-ahead logs, and retain free space inside pages for updates. Object stores charge for the object and for a minimum billable size that can be far above a small file. Filesystems allocate whole blocks, so a directory of 3 kB thumbnails occupies more than the sum of their lengths.
Then everything is multiplied again by whatever redundancy exists. Three-way replication triples the total; erasure coding is cheaper but not free; a backup policy with weekly fulls and daily increments can hold several more copies of most of it. A per-item figure of 40 kB that turns into 120 kB of billed storage is unremarkable, and it is a factor of three on an estimate that people generally argue about in fractions.
The estimate meets the two-systems problem at the end rather than the beginning. A drive, an array or a cloud bucket is sold in decimal terabytes — a trillion bytes each — while an operating system or a filesystem tool may report the same space in tebibytes of 1,099,511,627,776 bytes and print TB beside it. Ten per cent separates them, which at this scale is a hundred gigabytes per terabyte.
The input side carries a smaller version of the same problem. A per-item size taken from a tool that divides by 1,024 is really in kibibytes and understates the decimal figure by 2.4%. The two errors do not cancel; one is on the requirement and the other on the capacity, and a plan that is 10% over on one side and 2.4% under on the other is 12% short of where it thought it was.
Work in groups of three zeros and name the rungs as you go. A thousand kilobyte items is a megabyte, a million is a gigabyte, a billion is a terabyte. So a hundred million records at 10 kB each is a terabyte, and ten million images at 100 kB each is the same terabyte. Rehearsing two or three of those pairs makes most estimates answerable without a calculator.
The useful inversion is per-terabyte capacity rather than total size. A terabyte holds ten million 100 kB files, or a billion 1 kB records, or two hundred million 5 kB rows. Framing the question as "how many of these fit in a terabyte" catches errors that the forward direction hides, because the answer is a count somebody can compare against how many things they actually expect to have.
Most estimates that reach for terabytes should not have. If the answer comes out below one, the unit is wrong: 0.04 TB is a bad way of writing 40 GB, and the leading zeros hide the magnitude rather than conveying it. The threshold is not arithmetic but readability — quote in the unit that puts the significant digits to the left of the decimal point.
There is one case where the small decimal is the right answer, and that is when the terabyte is the unit being bought. A quota, a drive or a plan priced per terabyte makes 0.04 TB a meaningful statement about how much of the purchase is used. Outside that, an estimate that lands in gigabytes should be quoted in gigabytes and converted only if somebody asks how many of them fit on a drive.
One thousand million — 1,000,000,000 kB — because a terabyte is a trillion bytes and a kilobyte is a thousand. Turned around, that means ten million files of 100 kB fill a terabyte, or a billion records of one kilobyte each. Those two sentences are usually more useful than the conversion itself.
Arithmetically yes, editorially rarely. A figure that was worth measuring in kilobytes is nine orders of magnitude away from a figure worth quoting in terabytes, so the answer is either a very small decimal or the output of a multiplication that happened before the conversion. If it is the second, the estimate deserves stating as a multiplication rather than as a single converted number.
One, at most, and often none. When the input is an average per-item size that was itself rounded, the terabyte figure inherits that uncertainty multiplied by the count. Quoting 4.3271 TB from an estimate of "about 40 kB a record" claims a precision the 40 never had, and the extra digits make an assumption look like a measurement.
Indexes, which on a heavily queried table can equal or exceed the data. Replication, which multiplies everything by two or three. Block and page overhead, which rounds every small record up. Write-ahead logs, snapshots and free space held for reuse. A raw per-row figure is often half of what the storage bill is computed from.
A thousand, and the terabyte a trillion, both decimal. If the per-item size came from a tool that divided by 1,024, the input is really kibibytes and the estimate is 2.4% low; if the capacity it will be compared against is reported in binary units under decimal labels, that side is 10% out. Over nine orders of magnitude those are small errors, but they run in opposite directions.
Count zeros in threes. A thousand items of a kilobyte each is a megabyte, a million is a gigabyte, a billion is a terabyte. If a design involves fewer than a billion kilobyte-sized things it does not need terabytes, and if an estimate says otherwise the count or the per-item size has an extra group of three digits in it.
One TB is 1000000000 kB. It is the same relationship read backwards, so an answer from one page put through the other has to come back to where it started.
The claims this page makes about data units are checkable, and these are the documents that settle them.
The factor is a constant in the page and the arithmetic is four operations, so nothing is sent anywhere and nothing needs to be. The number you type never leaves the browser — there is no request for it to travel in.