When you think about it, punch cards are the safest way to backup data for long term storage.

They are not influenced by magnetic fields and their data (the holes) don't fade over time. They are also less sensitive to heat than most backup media I can think of.

Their biggest advantage should be that they are future-proof. Try reading an old computer backup tape or diskette (pick any size). Even if the data signal hasn't faded you probably cannot retrieve the data without some major effort.

Punch cards are easy enough to read and don't have proprietary encoding issues.

If you wanted to read your grandfather's punch cards (or: your grandson wanted to read yours) it should be easy enough to create a card reader.

It may not be ideal for graphics, but data - e.g. a family tree or government records - would be much safer if stored on punch cards in a fire-proof box.

How come punch cards are not used (or not very popular) for long-term backups?

  • 21
    Punch cards are easy enough to read and don't have proprietary encoding issues I don't know about you, but I can't just pick up a punch card and read it. What do the holes mean? Are they bits? Are the bytes in rows or columns? Which side of the card is up? If you choose a specific kind of card, like the aforementioned IBM-style, then you've decided on an encoding. That doesn't mean there isn't a proprietary encoding. The encoding is simple, but that's exactly because of the drawback: the data is very small and simple. Commented Feb 6, 2018 at 7:40
  • 4
    Ken Shirriff just wrote about restoring a 1402 — this goes into some detail on the inner workings of an old card reader. (@Raffzahn I think you’ll enjoy it, assuming you haven’t already seen it!) Commented Feb 7, 2018 at 18:02
  • 1
    There's a section about reading punch cards from the 1940s by a 1990s computer in Neal Stephenson's "Cryptonomicon". Commented Feb 8, 2018 at 5:45
  • 3
    They are ... less sensitive to heat Unless that heat gets up to Fahrenheit 451. Commented Feb 8, 2018 at 10:05
  • 1
    They question in the title makes an assumption that is not true. 5 years later and we still use punchcards here. Not for backup, but for parking keys. They are easy to mass produce, cheap and can hold enough information to store when you entered the parking garage. Commented Sep 29, 2023 at 15:26

13 Answers 13


The main reason punched cards aren't used any longer is density. A one-inch stack of cards is only 142 80-byte records (assuming the usual practice of encoding one byte per column). So if you need to store 50,000 records that's a stack of cards 350 inches tall (over 29 feet). And 50,000 80-byte records isn't that much, just 4 megabytes of data. If you were to take those 50,000 records and simply print them on paper it would take 758 pages (printed on both sides that's less than one 500-sheet package of paper). If you are paying for storage space (a warehouse, a floor of an office building, etc.) then consider the cost advantage of being able to store more records in less space.

By comparison, a relatively minor (but still important) reason is that modern records no longer fit onto a single 80-byte card. This means you need to use sequences of cards to hold larger records, which introduces a new problem:

Punched cards have no 'natural' sequence other than the sequence of cards in a stack. If you drop the stack or just insert one card in the wrong place the data may not be understandable unless you have a way to restore the original sequence. One way to record the sequence is to mark the sequence number on each card with a pen. This is time-consuming for a large stack of cards. Another way to record the sequence is to reserve some of the columns on the card to record a sequence number. The cards can then be sorted into the correct order mechanically (there are machines that will do this). The problem with doing this is that every column you reserve for sequencing reduces the number of columns you have left for actual data. So 50,000 cards might require 4 columns for sequence numbers, but now each card only holds 76 bytes of data instead of 80, so instead of 50,000 cards you need 52,632 cards, which makes your already-bad density problem even worse.

There are other ways to break up large records into multiple cards that don't involve sequence numbers, but they have similar problems that also end up requiring you to use more cards.

  • 25
    A classic way of saving card order was to take your stack and draw a diagonal line across the top with a fat marker. If dropped the cards could be sorted with a bit of effort back into the same order by reconstituting the diagonal stripe. Commented Feb 6, 2018 at 9:31
  • 4
    The classic way was to have organize your data in a way that sort order comes by nature - like customer and product ids. For programm data, every card had a sequence number. After all, punch card processing is build arround the capabilities of sorting machines. Not computers. Data processing before 'modern' memory monsters was a constant series of sort and merge runs with stages of counting and punching inbetween, toward a lister at the end (and new stacks).
    – Raffzahn
    Commented Feb 6, 2018 at 12:49
  • 1
    Common practice on important data was to limit the data to columns 1 to 72 and to use columns 73 to 80 as a sequence number, which allows the cards to be sorted in a card sorter to recover from them.
    – rcgldr
    Commented Feb 7, 2018 at 19:53
  • 4
    @rcgldr -- FORTRAN had a 72-character limit for exactly this reason. Ah, the good old days. They were terrible. Commented Feb 7, 2018 at 19:59
  • 1
    @Bitbang3r not even that. Use a blank punch card and print a few QR codes on top of it (that card is unpleasantly non-square). You have already increased the original data density, and you can use the spare space to punch out the card ID. If you use a decent scanner instead of a poorly focused camera (like my old phone had), you should even be able to use 40-L QR codes at 3kB per code. Commented Feb 9, 2018 at 10:08

Nice Question :))

Short answer: Density - It just takes way too much cards to store anything useful.

(And no, there is no way back in the good old time of optimized data structures)

When you think about it, punch cards are the safest way to backup data for long term storage.

As usual that depends on your definition on 'safest way'.

They are not influenced by magnetic fields and their data (the holes) don't fade over time. They are also less sensitive to heat than most backup media I can think of.

Like any other media they do need a careful storage. But you're right, they can endure quite some abuse and neglect.

Their biggest advantage should be that they are future-proof. Try reading an old computer backup tape or diskette (pick any size). Even if the data signal hasn't faded you probably cannot retrieve the data without some major effort.

Not true for classic nine track Tapes. they can not only be read by very primitive means (ferro fluid and a set of glasses), but are also (usually) plain text data. Of course, every data on every media may be application specific, and thus hard to decode for future data archeologists.

Punch cards are easy enough to read and don't have proprietary encoding issues.

Sure they have. Just because one physical storage is the same for all usages (size of cards and holes) doesn't mean all encodings are the same. While regular data may be encoded using zero to three holes per column to encode one symbols, there have been not only extended encodings with up to six holes per column for binary data. But even within just one supplier like IBM there were many ways to use the card. Including horizontal storage, where data words where stored in rows, not columns. And then there were other manufacturers. All using the same paper and hole dimensions, but completely incompatible encoding.

If you wanted to read your grandfather's punch cards (or: your grandson wanted to read yours) it should be easy enough to create a card reader. It may not be ideal for graphics, but data - e.g. a family tree or government records - would be much safer if stored on punch cards in a fire-proof box.

Or even use metal cards, stretching the durability even further.

How come punch cards are not used (or not very popular) for long-term backups?

The main reason is density - or it's reciprocal value size - punch dards carry only a minuscle amount of information compared to their size.

A punch card is 7.375 by ​3.25 by 0.007 inches (187.325 mm x 82.55 mm x 0,18 mm). That's a volume of ~2.8 cm³. With 80 bytes stored per card (*1), this comes down to 0.035 cm³ per character.

Thanks to the metric system everything from here only involve shifting the magnitue (*2).

  • One KiB is like 36 cm³ - like a small wallet
  • One MiB is 36,700 cm³ - already the size of a suit case (~10 US gallons)
  • One GiB is 37.58 m³ - that's already almost the size of a garage.
  • One TiB is 38,500 m³ - That's the size of a small school sports hall.

All of this without taking boxes and other measues to handle them.

Now tell me again, what's the size of your actual hard drive(s) 1 TB? 4 TB ?

*1 - We assume a simple standard code, no optimization that may give future readers a hard time.

*2 - I still used a pocket calculator, despite the fact that its a rough estimate anyway.


Then there is paper tape. Their density is about 10+ times of punch cards. Sure, a bit more fragile, but not so much. Still, even with 10 times less space needed, it can't be really used for today's surge of data.

Addendum 2:

Regarding the title ("Why no punch card readers in use?" ): They are ... at least I got one sitting right behind me, fully functional and in use.

  • 3
    For those values, are you confusing , as decimal separator and , as thousands separator?
    – wizzwizz4
    Commented Feb 6, 2018 at 20:46
  • 1
    Mylar is an alternative to paper for punched tapes. I recall companies in the 1970's and 1980's using mylar punched tape as backup / archive.
    – rcgldr
    Commented Feb 8, 2018 at 1:17
  • @wizzwizz4 the world is pretty well split on which punctuation should be used as the decimal / thousands separator.
    – Aaron
    Commented Feb 9, 2018 at 14:56
  • 1
    @Aaron But using the same punctuation mark for both is not very clear.
    – wizzwizz4
    Commented Feb 9, 2018 at 16:34
  • 1
    @Raffzahn disregard. I misread something above.
    – cbmeeks
    Commented Dec 21, 2018 at 14:11

Your premise fails because it presumes higher density methods can't also be used.

Back in the 1980s, the magazines that published code started using a barcode scheme so you could wand the program into your computer instead of keyboarding. It didn't take off, but the concept is sound. Scanning those magazine pages is trivial; Google Books has already done it. Writing some code to convert the scanned data into code again wouldn't be hard.

So you can either use one of those old methods rendered on microfiche as I'm sure has already been done... or you can design a new scheme for fiche, RAIDed so if a sheet is lost, or a corner or edge of all sheets a lost, the data can be reassembled.

And then, with a little industrial effort, you can sustain suport for any legacy format you please.

You are also placing too much faith in physical data rendering which itself has disadvantages. For instance you just change the problem from maintaining machines which can read it, to protecting a more balky and fragile physical media i.e. Paper. That tells me you are not an archivist.

And even this is predicated on the idea of sticking data on a shelf, ignoring it for 20 years and having it readable when you come back. That is an inept way to manage data that deserves to fail.

You protect data the way you protect the human genome: by continuously reproducing it.

  • 1
    Not realy an answer, but great touching the issue behind. Just one point, maintaining a bunch of paper (and implying reconstruction of a reader in some future) is way less effort than keeping a any computer alive for a long time. Not to mention less durable media.
    – Raffzahn
    Commented Feb 5, 2018 at 21:04
  • 4
    I know, unfortunately SE's policy of "always answer the question" fails when "the question is broke". Really it depends on the scale; archiving data on paper on a national scale is actually far more expensive than keeping a digital format alive. Cheaper still is to pay a monkey to stuff floppies once, and carry it forward on DASD from there. Commented Feb 5, 2018 at 21:17
  • 3
    I wouldn't call the question 'broke'. It is a valid one. That there is maybe more to think about on related issues ,does not dimish the value of the question asked.
    – Raffzahn
    Commented Feb 5, 2018 at 21:33
  • 1
    @Harper If you can't answer a question on stack exchange you can leave a comment explaining why, or refrain from answering. Stack Exchange is specifically designed to prevent "I just want to leave my 2 cents" type of answers.
    – pipe
    Commented Feb 7, 2018 at 8:43
  • 3
    @pipe that's pretty harsh. I did a fine job answeing the question, I just recognized a flaw in the underlying presumptions and addressed that too. As such, I did not simply retread/plagiarize/clone other answers. I am the only person talking about the economics of bulk storage vs maintaining legacy formats, because even hollerith cards don't read themselves. Commented Feb 7, 2018 at 17:45

Many years ago I worked in a shop that had desktop apps (VB6) and mainframes. I developed an app that replaced punch cards with ASCII text. This was in the 1998-1999 timeframe.

The biggest reason given was the card reader was highly mechanical and broke down several times a week. Eventually, parts were difficult (if almost impossible) to obtain. The business made a decision to replace with a more dependable technology.


There are also some failure modes you have overlooked. The GB academic community's copy of the 1966 census was stored on punch cards but when someone tried to reload it they discovered that a squirrel had nested in the punch cards and many were now unreadable.


The reason we don't use punch cards anymore is because literally every other data storage medium that's been invented since is vastly superior in every way imaginable. Let's compare them to the current media of choice for long-term data storage: magnetic tapes:

The first is obviously density, as every other answer here reflects, so I won't get into that much except to say that even if you used smaller holes and used an advanced encoding scheme with fewer holes needed to represent data, it would still be impossible by electro-mechanical means to achieve the density of literally billions of bits per square inch that magnetic media can provide.

The second concern is cost. Right now, a 7/15-terabyte LTO Ultrium tape can be had for around $80 -- almost a quarter of the cost of a hard drive with similar capacity. Assuming ultra high-density punch cards (that don't exist in reality), tapes would still be cheaper than paper.

Thirdly is fragility. Paper is way more fragile than plastic (which is what the backing material of tapes are made from). For long-term storage of punch cards, you need tightly controlled heat and humidity conditions for long periods of time. Paper molds, degrades, and warps much faster than plastic. If you made the punch cards out of plastic or even metal plates, they'd be thicker than paper and would contribute to the density problem above. Yes, magnetic tapes are sensitive to magnetic fields (obviously), but they're way less sensitive than you'd think, and modern error-correction algorithms make them very robust.

Lastly, let's not discount optical storage here either. Technically speaking, you can think of optical discs as a logical evolution of punch cards; the data is encoded using microscopic pits (i.e. "holes") in a hardy substrate that can be read by a laser rather than mechanical switches. Optical discs can therefore be made out of all sorts of materials, negating all the environmental concerns of punch cards and magnetic media. In fact, for serious long-term archival of certain vital records, optical discs are preferred in some cases even over tapes. Their only drawback is they still can't rival the density and cost advantages of other forms of media.

As for future-proofing (e.g. looking at your grandfather's porn collection), you could use vinyl records for that. The technology has been around for over a century, and it's been used before to encode digital data. It never really caught on, but there's a metal record floating around in space right now on the Voyager II probe, which aliens might find in a few thousand millennia. Doesn't get much more long-term than that.

Even though data storage on vinyl records has never been a thing, it'd still be a better choice than punch cards.

  • Apart from that parenthesised comment in the second-to-last paragraph, this is a good answer. +1, and welcome to Retrocomputing.
    – wizzwizz4
    Commented Feb 7, 2018 at 19:35
  • "7/15-terabyte LTO Ultrium tape can be had for around $80" - I can find LTO-7 tapes (6 TB native, ~15 TB compressed) for about $70-75.
    – bwDraco
    Commented Feb 7, 2018 at 19:44
  • 1
    Other media aren't better in every way imaginable. Punched cards have a major advantage that other media do not--which is also a major disadvantage that you neglected to mention: the ability to edit a stack by hand through the insertion and removal of cards [which unfortunately goes along with the ability to randomly edit the sequence by dropping a box of cards].
    – supercat
    Commented Dec 20, 2018 at 16:00

It may not be ideal for graphics, but data - e.g. a family tree or government records - would be much safer if stored on punch cards in a fire-proof box.

As already said, one card holds about 80 bytes of data, is made of paper, has the size of about, well, a postcard, and needs relatively low technology to read (or even, with some effort, you can train people to read it visually).

However, I can direct your attention to the wonderful retro technology of storing information on a roughly postcard-sized paper cards, requiring even less technology to store and retrieve the data (or even, with some effort, you can train people to read it visually): quotation slips. They were extensively used in lexicography, and I've indeed seen it in productive use (i.e. not only consulting existing old archives) only a few years ago.

And the storage density is substantially higher, my guess is that you can comfortably pack some hundreds of bytes per slip. Thus, when you already go this way, the quotation slips win over punchcards.


There are plenty of good reasons but to me the primary reason is the ease of copying. Two copies on bad mediums are more desirable than one copy on a good medium. I put a 4TB drive in a box and in less than a day have a complete copy on the medium of my choice. Don't try this with your box of punch cards at home.

Punch cards are only as good as the staff you're willing to pay to maintain them. I have bunches of hard drives and they don't require any staff.

  • Welcome to Retrocomputing Stack Exchange. Thanks for the answer; the second paragraph is a key insight. You might find the tour interesting.
    – wizzwizz4
    Commented Feb 7, 2018 at 20:10

Unfortunately, as someone who was around in those days of punch cards, a major con was the sheer size and costs of the card puncher and the reader. And as they were paper cards there was always the issue of card-jams/ripped cards, not only possible in the card puncher, but in the reader as well. A punch machine and reader/sorter required an area of about 10'x 5' and in those days business machines such as those and even typewriters were solid - built to last a hundred years. The only 'pro' was the bag of punchouts Mom brought home for us kids - I guess that would be a 'con' for the parents.

Also, if you had anything close to a sizable business (whatever that means) you would be spending a fortune on storage costs. My uncle ran a small mom & pop -sized accounting firm and their storage room was as big as the office itself, stacked floor-to-ceiling with cases of cards.

Oh, and did I mention the paper. You could lose your entire database in the event of a fire or water damage. And all of the paper dust (as well as those punchouts) that could easily ignite in those days when people smoked in their offices. Ew.


Other answers have focussed on data density and factors relating to the medium. However they are all missing one key element.

Individually robust artifacts do not give robust data security.

Data security does not come from having long-lived artifacts. We have a few scrolls which have survived from thousands of years ago - but most of them have long since crumbled to dust, or been burnt for kindling. We have some inscribed stones from thousands of years ago - but most of them have long since been reused for building material. Numerous artifacts have writing which cannot be decoded, because the language or script used has been lost to history, or which took centuries to be decoded until that language or script could be figured out. Ozymandias is the default state of archaeology.

What data has survived best from thousands of years ago? Answer: anything which has been copied multiple times, with multiple copies across multiple geographic locations. And of that, the data which has survived the absolute best is that which has not only been copied, but which has been repeatedly retranslated into the currently-applicable language. Anything with very few copies has generally been lost. The destruction of the Library of Alexandria, Nazi book burnings, or anything similar are all demonstrations that relying on a single storage location is not a robust strategy.

The Torah is perhaps the best example of how to preserve data. How robust is any single Torah? Not very - it's only paper. How robust is the data? History shows that it's been incredibly robust, because the process of copying a Torah does not just involve copying but also error-checking and error-correction. Checking between widely-separated Jewish communities, the error rate is tiny.

So, your punch cards. Yes, your single stack of punch cards could survive for a long time. As one fictional example shows, you could even use gold punch-cards. But unless you have multiple redundant copies, it's very likely that your single golden copy will succumb to damage and be lost; or that the necessary language/technology/whatever to read that golden copy will make it impossible to retrieve the data.


Using punchcard for long-term backup is a bad idea imho.

  1. density/storage space required - as pointed out in the other answers; 1 TB of data translate to 12.5 billion cards or (2.4 g per card?) 30,000 tons of cardboard
  2. sensitivity - paper easily gets moldy, brittle or warps; it's rather hard to store long term, especially in quantity
  3. pest - rats and mice eat paper as do various insects
  4. read speed - even with a high-speed mercury reader, reading back data is as slow as a nail; 30 cards per second (~2000 CPM) translate to 2,400 byte/s - that's little more than the infamous 1541 floppy drive. The 12.5 billion cards from above would take a good 13 years. yaaawwwwnnnnn...

It all depends on what "long term" means. 10 years? 100 years? 1000 years? You should carve your data in stone and bury it. ;-)


If one wants to use paper as a backup medium, there is always the option of using an OCR friendly font and printing the data - in plaintext or using a binary-safe encoding like base64. You will still get much better density. Or even printing it in a non-character, more compressed format.

Paper is not necessarily a more durable medium than eg a floppy disk or tape - the latter actually has less chance of catching mold or just weakening to a point where automated reading would tear it up when there is moisture or even flooding; silverfish and other paper eating critters will usually leave it alone; the risk of there being chemical self-decomposition is there with both media and well known for non archival grade paper.


An additional reason, not previously covered: expense.

The last anecdotally quoted price I can find for punch cards was from 1996, and was around USD 42 for a box of 2000 cards. That's a lot to pay for roughly 160 KB of storage.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .