Google, Amazon, Apple and 19 more: 22 founding legends traced to their sources

Open Amazon’s 1997 IPO prospectus and Jeff Bezos’s parents’ investment reads like this. In February 1995, 582,528 shares went to his father. In July of the same year, 847,716 shares went to a trust whose beneficiary was his mother, both at $0.1717 a share. Together that comes to about $246,000. The same prospectus even lists, in three entries, the amounts Bezos himself lent the company interest-free: $15,000 in July 1994, $29,000 in November and $40,000 in November 1995. Numbers that founding stories almost never mention sat on paper, each with a date.

What is left if the numbers of legendary founding years are traced back one by one to their originals? To find out, 22 companies were lined up.

There were only two criteria for selection: the company’s founding story is known across borders, and at least one of its founding-era numbers could be confirmed in a source free to read. The list centres on Japanese and American companies, with founding years scattered from 1889 to the late 2000s. Well-known companies whose founding numbers could not be traced to a readable source were left out.

22 companies: the numbers confirmed on paper and the scenes that get retold

The middle column and the two on the right describe the same founding years, but the character of their sources differs. The middle holds capital, registrations, share counts, revenue and contract terms, published by the company, a regulator or a buyer. The two on the right hold scenes of what people did and what happened, published by the founders themselves or by people who interviewed them.

The plain numbers were written by people with a reason to publish

The numbers on paper carry the circumstances of whoever published them. The share counts and revenue in a prospectus are statutory disclosures made on listing. eBay and Google wrote down acquisition prices and exchange ratios because a listed company buying a business takes on disclosure obligations. The histories of HP, Nidec and Nintendo are what those companies keep as their own beginnings.

Failures were mixed in among them. Uniqlo’s official history records that the first store, in Hiroshima, closed in 1991. Falcon 1’s three failures remain intact as launch records. Both, though, are single lines of fact rather than scenes. Rocket launches happen in public, and a closed store vanishes from the map. Only failures that could not be hidden made it onto the paper side.

Seven of eight scenes from the hard years were not on paper

Pull the scenes of hard times out of the right-hand column and there are eight: Panasonic’s socket, Sony’s rice cooker, Rakuten’s first month, Netflix and Google being turned down by buyers, Musk borrowing for rent, YouTube’s dating site that nobody came to, and Airbnb’s cereal boxes.

Only one of these, Panasonic’s socket, appeared in a company document, and even there the founder’s own memoir sits alongside the company history as a source. Five, from Netflix, Google, Musk, YouTube and Airbnb, were told after success by the founder or a party to events, or written into a book by someone who listened. The remaining two, Sony’s rice cooker and Rakuten’s 320,000 yen, are widely told but could not be confirmed in readable sources.

When they were told is skewed as well. The Google story came 11 and 15 years later, YouTube’s in a speech the year after the sale, Netflix’s in a book 19 years on, and Zip2’s starting money in a biography after 20 years and the founder’s own post after 24. Every one surfaced once the company’s success could be assumed.

A failure becomes legend only when someone who succeeded tells it. The desk reads the scarcity of failure scenes on paper not as concealment but as a sign that, at the time, nobody had a reason to write them down.

Each retelling nudges the numbers a little

The numbers in the scenes wavered more as narrators multiplied.

For Microsoft’s first revenue, Wikipedia gives $16,005 through the end of 1976 (with no source shown in the text), while a computing history museum’s material puts 1975 gross income at $1M. That is two orders of magnitude apart. Airbnb’s cereal-box revenue moves between $20,000 and $30,000 depending on which book is the source, and Wikipedia simply gives the range. Zip2’s starting money splits into $15,000 and $28,000, and the reason Google was turned down splits into price and terms.

A number on paper has only one version. A remembered number has as many versions as it has narrators.

The failures in the cases recorded here were written before success

On the far side of this table sits the opposite way of keeping records. The maker of Genji Reincarnation wrote, one month after release, that a game built over about three years had sold 32 copies. Promotion cost about 20,000 yen, and the ceiling on revenue fell below it. The record of two years of LINE stickers for 1,807 yen in total, the record of a food truck earning 400 yen an hour in its first month and the record of an AI explainer channel at 416 yen an hour after two years were all written not after success but along the way, or without success ever arriving.

The failures on the legend side are told only once success provides an ending. These failures have no ending, and when they were written the person writing was still inside them. As numbers, they sit closer to the original.

The limits of this count

Twenty-two is small against the full population of legendary companies, and the selection depended on whether readable sources existed. For Sony and Panasonic, whose histories could not be read directly from the official sites, the count relied on third-party summaries based on company histories. Where the lines fall is the desk’s own judgment. Panasonic, for example, with its company history and the founder’s memoir side by side, is counted on the “in a company document” side. A scene marked unconfirmed does not mean it never happened, only that its original could not be reached within what is free to read.

The founding years of Google and of Musk each got their own deeper look: the story of Google nearly selling for $750,000 and the starting money and sale prices of Musk’s first company. The same lean in where numbers come from showed up in the twelve-year timeline of Chiikawa.

When reading a legend’s numbers, look first at who wrote them

After lining up all 22, only the right-hand column stayed on the desk. Nearly every scene remembered as legend was told by someone who had been at the company, after it had grown. That is not to call them false. They are scenes chosen in hindsight by people who knew about the success, not records written on the day.

So when reading the numbers of a legend, the rule here is to look at who wrote them and in what year before looking at the numbers themselves. The company and the authorities write the capital and the share counts, and the founder who succeeded writes the failure scenes later. The 32 copies were written one month after release and the 1,807 yen after two years, both by people who did not yet know how it would end.