The History of Startups: Where It Came From and Where It Goes

— I'll answer that, I promise. But first let's work out what exactly appeared there. Otherwise we'll end up talking about where the tree grew without a single word about what kind of tree it was. Name me a startup that called itself a startup in 1957.

— Nobody. Not one. The word in its current sense simply didn't exist then. The Oxford English Dictionary gives the earliest use as a 1976 Forbes note about an unfashionable way of putting money into startups in electronic data processing. The phrase "startup company" is recorded a year later still, in Business Week. Which means HP in 1939, Fairchild in 1957 and Intel in 1968 physically could not have called themselves startups. They called themselves a new company. A small business. Sometimes just "us."

— Exactly — in essence. And what is that essence? Let's test your criteria. Size? HP had two people in a garage. A company that today writes "we're a startup" in its job postings can have a thousand. Industry? In its first years HP made a foul-line indicator for bowling, a device that flushed a urinal by itself when a man walked up to it, and a weight-loss machine that ran current through you. Hewlett later recalled that they'd take on anything for a nickel. The starting money was 538 dollars. The success turned out to be the fourth or fifth attempt — an audio oscillator, eight of which were bought by the Disney studio to equip theaters for "Fantasia."

— What's left is a state. In 2010 Steve Blank formulated a definition that became the working standard in business schools: a startup is a temporary organization built to search for a repeatable and scalable model, not to execute one.

— By that definition, yes. And it's the only definition with an ending built into it. Which implies something fairly insulting, by the way: the line "we've been a startup for 8 years now" isn't pride, it's a diagnosis. Either you found the model and became a company, or you didn't find it and shut down. There's no third option in the definition.

— It's searching. Just for something else. The barber is looking for a cheap lease and a way to lure people in from the next street over, but the model by which he'll make money has been known for about 100 years: a haircut costs this much, rent costs that much, the difference is yours. He isn't testing a hypothesis, he's executing a scheme. Paul Graham puts it even more bluntly: a startup is a company designed to grow fast, and a barbershop simply isn't designed that way. In his words, neither being newly founded, nor working with technology, nor taking venture money, nor selling the company makes it a startup on its own. The only essential thing is growth.

— As is. There's a third — Eric Ries: a human institution creating a new product under conditions of extreme uncertainty. He deliberately said nothing about size or industry, because he wanted it to work inside a huge corporation too.

— It has one. Three, in fact, and they don't agree. Take a company that found its model and is growing 30% a year, steadily and profitably. By Blank it's no longer a startup. By Graham it isn't either — 30% a year isn't the pace; he's talking about 5–7% a week. And a team inside Google launching a new product in total fog is a startup by Ries — even though it isn't a separate company and takes no venture money at all.

— None. These are definitions, not discoveries. They're chosen, not found in nature. I'll tell you honestly: the idea that a startup is a mode rather than a type of company is the position of the three most-cited authors on the subject, not an established fact. It can't be proven, in principle. But it's useful, because it makes it immediately visible where the word came loose from the meaning.

— WeWork. On August 14, 2019 they file to go public. The first sentence of the document says they are a community company committed to maximum global impact. The implied valuation is around 47 billion dollars. Six weeks later the filing is withdrawn, and the valuation being discussed by then is already below 10.

— By Graham, unquestionably: it grew fast. By Blank, no: the model was never found. The difference between the two definitions came out to 37 billion dollars, and that, to my mind, is the best argument for understanding the word precisely.

— So much so that a new word had to be invented for rarity. At the end of 2013 Aileen Lee called 39 American software and internet companies unicorns — founded since 2003 and valued above a billion. And she also counted the denominator: roughly 60 thousand venture-funded companies over that same decade. She herself noted that 60 thousand was a rough extrapolation from three different estimates, so it's an order of magnitude, not a census. But the order is this: 0.07%. About one in fifteen hundred.

— Because nobody writes about the fifteen hundred. You know the name of one. And now we can come back to your question about place. If a startup is a mode of searching, then the question "why exactly there" stops being a question about climate and smart people. It becomes a question about where it's cheapest to be wrong many times in a row.