Y Combinator：创业生态的新物种

March 2008, rev May 2013(This essay grew out of something I wrote for myself to figure out what we do. Even though Y Combinator is now 3 years old, we're still trying to understand its implications.)

I was annoyed recently to read a description of Y Combinator that said "Y Combinator does seed funding for startups." What was especially annoying about it was that I wrote it. This doesn't really convey what we do. And the reason it's inaccurate is that, paradoxically, funding very early stage startups is not mainly about funding. Saying YC does seed funding for startups is a description in terms of earlier models. It's like calling a car a horseless carriage.

When you scale animals you can't just keep everything in proportion. For example, volume grows as the cube of linear dimension, but surface area only as the square. So as animals get bigger they have trouble radiating heat. That's why mice and rabbits are furry and elephants and hippos aren't. You can't make a mouse by scaling down an elephant.

YC represents a new, smaller kind of animal—so much smaller that all the rules are different.

Before us, most companies in the startup funding business were venture capital funds. VCs generally fund later stage companies than we do. And they supply so much money that, even though the other things they do may be very valuable, it's not that inaccurate to regard VCs as sources of money. Good VCs are "smart money," but they're still money.

All good investors supply a combination of money and help. But these scale differently, just as volume and surface area do. Late stage investors supply huge amounts of money and comparatively little help: when a company about to go public gets a mezzanine round of $50 million, the deal tends to be almost entirely about money. As you move earlier in the venture funding process, the ratio of help to money increases, because earlier stage companies have different needs. Early stage companies need less money because they're smaller and cheaper to run, but they need more help because life is so precarious for them. So when VCs do a series A round for, say, $2 million, they generally expect to offer a significant amount of help along with the money.

Y Combinator occupies the earliest end of the spectrum. We're at least one and generally two steps before VC funding. (Though some startups go straight from YC to VC, the most common trajectory is to do an angel round first.) And what happens at Y Combinator is as different from what happens in a series A round as a series A round is from a mezzanine financing.

At our end, money is almost a negligible factor. The startup usually consists of just the founders. Their living expenses are the company's main expense, and since most founders are under 30, their living expenses are low. But at this early stage companies need a lot of help. Practically every question is still unanswered. Some companies we've funded have been working on their software for a year or more, but others haven't decided what to work on, or even who the founders should be.

When PR people and journalists recount the histories of startups after they've become big, they always underestimate how uncertain things were at first. They're not being deliberately misleading. When you look at a company like Google, it's hard to imagine they could once have been small and helpless. Sure, at one point they were a just a couple guys in a garage—but even then their greatness was assured, and all they had to do was roll forward along the railroad tracks of destiny.

Far from it. A lot of startups with just as promising beginnings end up failing. Google has such momentum now that it would be hard for anyone to stop them. But all it would have taken in the beginning would have been for two Google employees to focus on the wrong things for six months, and the company could have died.

We know, because we've been there, just how vulnerable startups are in the earliest phases. Curiously enough, that's why founders tend to get so rich from them. Reward is always proportionate to risk, and very early stage startups are insanely risky.

What we really do at Y Combinator is get startups launched straight. One of many metaphors you could use for YC is a steam catapult on an aircraft carrier. We get startups airborne. Barely airborne, but enough that they can accelerate fast.

When you're launching planes they have to be set up properly or you're just launching projectiles. They have to be pointed straight down the deck; the wings have to be trimmed properly; the engines have to be at full power; the pilot has to be ready. These are the kind of problems we deal with. After we fund startups we work closely with them for three months—so closely in fact that we insist they move to where we are. And what we do in those three months is make sure everything is set up for launch. If there are tensions between cofounders we help sort them out. We get all the paperwork set up properly so there are no nasty surprises later. If the founders aren't sure what to focus on first, we try to figure that out. If there is some obstacle right in front of them, we either try to remove it, or shift the startup sideways. The goal is to get every distraction out of the way so the founders can use that time to build (or finish building) something impressive. And then near the end of the three months we push the button on the steam catapult in the form of Demo Day, where the current group of startups present to pretty much every investor in Silicon Valley.

Launching companies isn't identical with launching products. Though we do spend a lot of time on launch strategies for products, there are some things that take too long to build for a startup to launch them before raising their next round of funding. Several of the most promising startups we've funded haven't launched their products yet, but are definitely launched as companies.

In the earliest stage, startups not only have more questions to answer, but they tend to be different kinds of questions. In later stage startups the questions are about deals, or hiring, or organization. In the earliest phase they tend to be about technology and design. What do you make? That's the first problem to solve. That's why our motto is "Make something people want." This is always a good thing for companies to do, but it's even more important early on, because it sets the bounds for every other question. Who you hire, how much money you raise, how you market yourself—they all depend on what you're making.

Because the early problems are so much about technology and design, you probably need to be hackers to do what we do. While some VCs have technical backgrounds, I don't know any who still write code. Their expertise is mostly in business—as it should be, because that's the kind of expertise you need in the phase between series A and (if you're lucky) IPO.

We're so different from VCs that we're really a different kind of animal.

Can we claim founders are better off as a result of this new type of venture firm? I'm pretty sure the answer is yes, because YC is an improved version of what happened to our startup, and our case was not atypical. We started Viaweb with $10,000 in seed money from our friend Julian. He was a lawyer and arranged all our paperwork, so we could just code. We spent three months building a version 1, which we then presented to investors to raise more money. Sounds familiar, doesn't it? But YC improves on that significantly. Julian knew a lot about law and business, but his advice ended there; he was not a startup guy. So we made some basic mistakes early on. And when we presented to investors, we presented to only 2, because that was all we knew. If we'd had our later selves to encourage and advise us, and Demo Day to present at, we would have been in much better shape. We probably could have raised money at 3 to 5 times the valuation we did.

If we take 7% of a company we fund, the founders only have to do 7.5% better in their next round of funding to end up net ahead. We certainly manage that.

So who is our 7% coming out of? If the founders end up net ahead it's not coming out of them. So is it coming out of later stage investors? Well, they do end up paying more. But I think they pay more because the company is actually more valuable. And later stage investors have no problem with that. The returns of a VC fund depend on the quality of the companies they invest in, not how cheaply they can buy stock in them.

If what we do is useful, why wasn't anyone doing it before? There are two answers to that. One is that people were doing it before, just haphazardly on a smaller scale. Before us, seed funding came primarily from individual angel investors. Larry and Sergey, for example, got their seed funding from Andy Bechtolsheim, one of the founders of Sun. And because he was a startup guy he probably gave them useful advice. But raising money from angel investors is a hit or miss thing. It's a sideline for most of them, so they only do a handful of deals a year and they don't spend a lot of time on the startups they invest in. And they're hard to reach, because they don't want random startups pestering them with business plans. The Google guys were lucky because they knew someone who knew Bechtolsheim. It generally takes a personal introduction with angels.

The other reason no one was doing quite what we do is that till recently it was a lot more expensive to start a startup. You'll notice we haven't funded any biotech startups. That's still expensive. But advancing technology has made web startups so cheap that you really can get a company airborne for $15,000. If you understand how to operate a steam catapult, at least.

So in effect what's happened is that a new ecological niche has opened up, and Y Combinator is the new kind of animal that has moved into it. We're not a replacement for venture capital funds. We occupy a new, adjacent niche. And conditions in our niche are really quite different. It's not just that the problems we face are different; the whole structure of the business is different. VCs are playing a zero-sum game. They're all competing for a slice of a fixed amount of "deal flow," and that explains a lot of their behavior. Whereas our m.o. is to create new deal flow, by encouraging hackers who would have gotten jobs to start their own startups instead. We compete more with employers than VCs.

It's not surprising something like this would happen. Most fields become more specialized—more articulated—as they develop, and startups are certainly an area in which there has been a lot of development over the past couple decades. The venture business in its present form is only about forty years old. It stands to reason it would evolve.

And it's natural that the new niche would at first be described, even by its inhabitants, in terms of the old one. But really Y Combinator is not in the startup funding business. Really we're more of a small, furry steam catapult.

Thanks to Trevor Blackwell, Jessica Livingston, and Robert Morris for reading drafts of this.

Comment on this essay.