“A world market for about five computers”
The usual version goes like this: in 1943 IBM’s head Thomas Watson said the world would need about five computers. It gets told as a textbook case of short-sightedness.
Nobody has found a document in which Watson says it. At the end of August 2026 Quote Investigator published an investigation and found no substantive evidence for either Watson Sr. or his son. IBM itself searched speeches, letters and meeting notes, and in a 2002 FAQ put the line down to a misunderstanding.
The misunderstanding looks like this. On 28 April 1953, at the stockholders’ meeting, Thomas Watson Jr. spoke about the IBM 701 and a trip “on which we expected to get orders for five machines, we came home with orders for 18”. The real sentence says the opposite of the myth: they expected five orders and got more than three times as many.
The year “1943” came later. The earliest attribution to Watson found so far is the 1981 book Facts and Fallacies, with no precise reference. The date was added by The Experts Speak in 1984, citing that same Facts and Fallacies. Early computer builders did hold similar expectations: Howard Aiken, as Popular Science Monthly reported in October 1944, could not imagine enough work for more than one such machine. But that is a different man and a different sentence.
“640K ought to be enough for anybody”
The story: in 1981 Bill Gates said 640 kilobytes of memory would be enough for anyone. Nobody has found a 1981 text where that is written down.
The earliest appearance Quote Investigator found is an editorial by James E. Fawcette in InfoWorld on 29 April 1985. It opens with words signed with Gates’s name, and the words are different: when the upper limit of PC-DOS was set at 640K, the thinking was that nobody would ever need that much memory. The editorial does not say where or when Gates said this, and the words are not from an interview.
Gates answered for himself. His column in the Daily News of Los Angeles on 22 January 1996 says: “I’ve said some stupid things and some wrong things, but not that.” In the same column he complains that the quote never comes with a citation and just keeps circulating, like a rumour. In 2001, in an interview with U.S. News & World Report, he repeated that he had said the opposite. By his account, the IBM PC was at one point going to be limited to 512K, and the limit had to be pushed up again and again. Computerworld retells that interview.
Forty-one years have passed since the InfoWorld piece. The line still has no primary source.
The horse Ford never asked for
Henry Ford died in 1947. The line about customers who, if asked, would have wanted a faster horse was first put in his mouth in 2001, in a reader’s letter to Marketing Week.
Before that, the idea lived without Ford. Quote Investigator found the precursors. In 1930 Edward Menge pointed out in The Quarterly Review of Biology that the steam locomotive did not come from breeding faster horses. Lewis Mumford made a similar point in 1946. In 1999 John McNeece, writing in a quarterly cruise-industry magazine, already named Ford, but only hypothetically: had Ford canvassed people, they would probably have told him they wanted a faster horse.
Two years later the hypothetical example became Ford’s own words, in the first person. The familiar plural, “faster horses”, turns up in 2003 in a book by Dick Lynch, and even there with the caveat that Ford “supposedly” said it. In mass circulation the caveat fell away.
That is how a quote gets assembled: first an idea with no author, then a hypothetical example with a famous name, then direct speech. Each retelling makes it smoother and strips out a little more doubt.
Einstein and the “generation of idiots”
Two lines about technology are pinned on Einstein. The first claims that our technology has appallingly outstripped our humanity. The second, usually over a picture of people staring at their phones: he supposedly feared the day technology would surpass human interaction, leaving the world with a generation of idiots.
In almost that form, the first line is spoken in the 1995 film Powder, written and directed by Victor Salva. The teacher Donald Ripley, played by Jeff Goldblum, says it, and the character Jeremy Reed credits it to Einstein. The line is not in The Ultimate Quotable Einstein from Princeton University Press, Quote Investigator notes.
The second meme was examined by the fact-checker Full Fact in July 2021. There is no evidence Einstein said it. The earliest versions appeared online in 2012. A similar line is in the same film, but the fact-checkers show no documented chain from the film to the meme.
Einstein does have a genuine line on the same theme, with a date. On 25 May 1946 The New York Times printed his words that the unleashed power of the atom had changed everything except the way we think. That is 66 years older than the first versions of the meme.
What was actually said
Real statements have an author, a date, a publication and a page. You can argue with them on the merits.
Gates, afterword to the revised 1996 edition of The Road Ahead, page 316: “People often overestimate what will happen in the next two years and underestimate what will happen in ten” (Quote Investigator). In 1999, in Business @ the Speed of Thought, he repeated the idea in almost the same words.
This is a prediction about misjudging the horizon, not about a particular machine. The wording is Gates’s; the idea is older. Quote Investigator traces it to Licklider (1965), Heilmeier (1976) and Roy Amara (1983). It is often called Amara’s law. The IBM 701 story only partly fits the pattern: in 1953 even near-term demand was underestimated.
Marc Andreessen, The Wall Street Journal, 20 August 2011, the essay Why Software Is Eating the World. It opens with the sentence “Software is eating the world.” The full text is reposted on the a16z site with a note on where and when it first ran. You can argue with this prediction because you can open it and read it again.
Older research papers pass the same test. Lisanne Bainbridge’s 1983 paper on the ironies of automation runs to five pages in the journal Automatica, and its conclusion that automation leaves people the hardest cases can be checked against the original.
We check quotes for our own articles the same way. Our piece on automating inefficient processes uses Gates’s rule that automating an inefficient operation multiplies the inefficiency. It is often repeated without a source. We found it in scans of two 1995 editions of The Road Ahead, at the end of chapter seven. Of the first sentence of the same rule, the scan shows only the beginning, so in that piece we paraphrase it and keep it out of quotation marks. By the same method we traced the lineage of the line “AI won’t replace you, a person using AI will”: the earliest version we found is a radiologist’s tweet from 2017.
How to check a quote in ten minutes
The order we work in:
- Find the earliest dated appearance. You need a specific text with a date; “widely quoted” does not count. If decades separate the supposed words from the first publication, that alone is a warning sign.
- Check the date against the biography. Ford died in 1947; the direct attribution appeared in 2001. Sometimes one subtraction is enough.
- Open the scan. Open Library’s full-text search across digitised books finds a phrase in a specific edition. That is how Gates’s line about two years and ten turns up in the 1996 editions.
- Treat an aggregator as a hint. A quotes site that gives the line a name but no book and no page tells you only that someone attributed it.
- Compare the wordings. If a line drifts between “about five” and “maybe five”, “horse” and “horses”, it was most likely retold from memory.
Numbers in AI sales decks get the same treatment. A savings percentage, a share of failed pilots, “three times faster”: the questions are the same. Who counted, on what sample, in what year, and does the document itself open, rather than a retelling of a retelling. If the chain ends at somebody else’s slide, the number has the same status as “640K”. When we break down someone’s process, figures from other vendors’ AI pitches go through exactly this check.
What we don’t know
A missing source does not prove a line was never said. The honest status of “640K” is “unconfirmed”, not “disproven”: Quote Investigator says only that it would not attribute the line to Gates at this time.
According to fact-checkers, The Henry Ford museum keeps a database of authenticated Ford quotations, and the “faster horse” is not in it. We opened neither those fact-checks nor the database, so our conclusion about Ford rests on Quote Investigator’s investigation. The “five computers” investigation is also a secondary source, from August 2026; we did not look at IBM’s archives. The link between the “generation of idiots” meme and the film Powder is plausible but not proven.
If your process has a number everyone believes but nobody has recounted, such as how many hours a week go on sorting incoming enquiries, describe the process in the short questionnaire. We break down one process for free, within 48 hours and without a call.
