Y Combinator: How a Startup Accelerator Changed Silicon Valley

March eleventh, two thousand five. Two people walking home from dinner in Harvard Square, arguing about everything that's broken with venture capital.

Arguing, or complaining? Because I do that walk home too and nothing comes of it.

Usually nothing does. But by the time Paul Graham and Jessica Livingston turned onto Walker Street, they'd decided to start their own investment firm. He put in a hundred thousand dollars. Robert Morris and Trevor Blackwell put in fifty each.

So two hundred thousand total. That's... a nice house, not a Silicon Valley institution.

And they almost called it Cambridge Seed. That name never saw daylight.

Why not?

They realized what they were doing could be national, and they didn't want a name tied to one town. So, Y Combinator.

Okay, but what was actually broken? VCs existed. Rich people existed.

That was the problem. Jessica's described it plainly: in two thousand four, if you wanted the first money for your startup, you found a rich relative or a friend's uncle. After that it jumped straight to VCs wanting to put in millions.

So nothing in between.

Nothing standardized. Graham points out Facebook started as a Florida LLC. That's how random the paperwork was.

A Florida LLC. That's a company that rents jet skis.

And Graham had lived the other version. When he started Viaweb, a friend named Julian Weber gave them ten thousand dollars, set up all the legal paperwork, taught them what business even was, and stayed calm when things went wrong. For that he got ten percent.

Ten percent for ten grand? That sounds brutal.

Graham thought the same thing. Then, he says, a second later he realized that without Julian, Viaweb never makes it. Good deal for Julian, good deal for them.

So they're trying to be everyone's Julian. How do you do that at scale?

By accident, mostly. Neither of them knew anything about angel investing. So Jessica said, let's learn fast, fund a bunch of startups at once, all in one summer.

Wait, that's the famous part. The batch. The whole model.

And they didn't know it was the important part. They thought they'd do one-at-a-time investing like everyone else. The summer thing was a training exercise.

They invented the thing by trying to practice the other thing.

Graham wrote a website, they put it up, and Jessica said they just hoped people would apply. Eight startups made it into that first summer. How many of the eight do you think worked out?

Eight companies, two thousand five, nobody's heard of them... I'll say one. Maybe one and a half.

Graham wrote in two thousand seven that at least four succeeded. One of them was a merger of two companies in that batch. Reddit and Infogami.

Reddit was in the FIRST eight? That's not luck, that's absurd.

Graham didn't think it would hold either. He said that batch could have been an anomaly and he'd feel safe aiming at twenty-five percent long term.

What about the ones that died? Because that's the part nobody tells you.

There's one called Kiko, a calendar company, squashed by Google Calendar after a year. They sold the software on eBay for a quarter of a million dollars, paid back their investors, walked away with about a year's salary each.

That's not a disaster, that's a soft landing.

And then they immediately started something else. Justin dot TV. Which, years later, Amazon bought as Twitch for about nine hundred and seventy million dollars.

So the failure from batch one became the billion-dollar thing. That's almost unfair.

Now watch the numbers move. Two thousand fifteen: about nine hundred and forty companies funded, eight worth over a billion.

And now?

By the end of twenty twenty-two, eighty companies valued over a billion, and the total value of YC companies past six hundred billion dollars. One batch alone, the twenty-twelve cohort, had Coinbase and Instacart in it. PitchBook put that single cohort's exit value at about ninety-three billion.

Against what, for everyone else?

Techstars' best was under ten billion. And YC's unicorn rate for its twenty-ten to twenty-fifteen batches was about five point eight percent, against two point two for Techstars.

Okay but at some point it stops being a cute summer camp and starts being a gatekeeper. Does that bother people?

It does. The deal now is five hundred thousand dollars, but the structure is seven percent for a hundred and twenty-five thousand, plus three hundred and seventy-five that converts later. Forbes pointed out that's not cheap, especially since all the teaching materials are free on the internet.

So you're paying seven percent for the phone numbers, basically.

For the signal. In one recent batch they picked four hundred and one companies out of more than sixteen thousand applicants. Getting in IS the product.

And one minute each on Demo Day to justify it.

One minute. Three hundred and seventy-seven companies pitching over two days.

What gets me is the first batch wasn't a strategy. It was two people who didn't know what they were doing, deciding to learn in public.

And Graham's answer to why more people don't try is that nearly everyone who does it prefers it to a regular job. He said the most successful founders are usually the uncertain ones. Some of them only applied hours before the deadline.

So the lesson is, apply anyway.

The lesson is they almost called it Cambridge Seed and went asynchronous like everybody else. Half of that company is just the version of it they didn't do.

I'm going to think about that on my next walk home.