Patrick Collison: "What If You Succeed?"
Stripe's CEO on schlep blindness, the question founders skip, and why new businesses are starting at twice last year's rate.
In 2009, Patrick and John Collison went to Startup School in Berkeley, got sushi in Potrero Hill afterward, and decided on the walk home to start Stripe. The reasoning, as Patrick remembers it, was that “we might as well because it probably won’t be that hard.”
It took them two years to launch.
Seventeen years later, at Startup School 2026, he talks with Harj Taggar about dropping out of MIT twice, why founders should ask what happens if they succeed, and what Stripe’s own data says about the best time to start a company.
Timestamps
00:07 — What Should You Still Learn in the Age of AI?
02:01 — Knowledge Still Matters
05:12 — Should You Drop Out of College?
09:58 — Why Stripe Worked
12:10 — Building Stripe Before Launching
17:20 — Is the Lean Startup Still the Right Playbook?
19:07 — The Hidden Reward of Building Stripe
22:36 — Will AI Kill Your Startup?
25:17 — Why It's Never Been a Better Time to Start a Company
29:23 — What Stripe's Data Says About the AI Economy
30:45 — Build Something People Truly Need
Transcript
Harj Taggar: Okay. Patrick, thanks so much for being here. Welcome to Startup School.
Patrick Collison: Great to be here. Harj and I first met 20 years ago and we started a company together. Sorry, am I giving away the introduction?
Harj: I thought this was my interview, but keep going. You’re doing a great job.
Patrick: I was going to say we started a company together many years ago and I learned a huge amount from Harj. So it’s really fun to do this too.
Harj: Right. Well, actually, speaking of that, I think when I first met you 20 something-ish years ago, at the time, your most impressive achievement I would argue was Croma, your dialect of Lisp.
Patrick: Any Lisp programmers here? Oh wow. Okay. I think I heard one whoop, which is more than I expected. But yeah, I really liked Lisp when I was in high school.
Harj: Yeah. So what I was going to ask is, a prolific 16-year-old today could presumably just prompt Claude to write their Lisp dialect. Would you advise them to not do that and still do it? Is there any value in such things?
Patrick: I don’t know. I wonder a lot. Obviously on the one hand, it used to be really fun to write all this assembly and machine code and to optimize your instructions and layout and memory and everything. And now we don’t have to do that anymore. Compilers do it for us. We don’t mourn it too much. And so maybe in the same way we shouldn’t mourn source code. We should just transcend the plane of instructions to Claude et al. But emotionally, I miss it.
Harj: How about, I think just as I’ve been hanging out here with these students, maybe the question behind it is many of them are just wondering what should they be learning at college? In this AI world, how much should they be trying to learn and derive from first principles and how much should they just outsource to the AI?
Patrick: Right. My model of this is cache, the CH, not an SH, where Jeff Dean has this famous set of numbers that every programmer should know—bandwidths and latencies and just kind of relevant constants that you should reason about as you build systems. And obviously when you’re thinking of building any system or distributed system or whatever, all lookups and all relevant bandwidths between different components are very different. Retrieving something from L1 cache is very different from retrieving from RAM, which is very different from retrieving across the network or whatever. And I think it’s like that with knowledge where, fine, yes, you can ask the agent or something to compute something for you or to look something up for you or whatever. That’s a hell of a lot slower than knowing it in cognitive L1 cache. And you can have way more round trips in your brain than you can muttering through Superwhisper or typing it out or whatever.
And so I think even granting the full capabilities of the models, I still think there’s a pretty— I think for a long time to come, neuronal lookups will be much faster. And then, if you look in revealed preference at what companies themselves are doing, whether they’re companies like Stripe or the labs or what have you, there still seems to be an enormous premium on cognitive ability. And so I wouldn’t— I think renouncing that before there’s evidence that we’ve saturated those benefits would be premature.
Harj: Are there specific things that maybe you personally, either personally or as CEO of Stripe, you still purposely choose to do yourself and retrieve from your own cache, even though the agents would probably do a reasonably good job?
Patrick: I still write myself. I both philosophically, but also specifically, substantively dislike the writing of the models. It’s very interesting, right? Because these can prove the Jacobian conjecture, whatever. And so clearly they’re capable of these monumental feats. But somehow I still haven’t read the LLM essay that I’ve found super compelling. It’s just very hard to RL limit that domain because the utility function or something is kind of hard to define. But yeah, I think writing is pretty—interpersonal communication and writing I think are so very fundamental and being able to reason sensibly in the multidimensional space of reality. And in some kind of indescribable way, I feel like the model is still kind of deficient at that. And so I’ve yet to send—every tool is now trying to prompt me with pre-written suggestions, whether it’s Gmail or apparently WhatsApp just rolled this out.
And I think I’ve still sent zero of those in my life.
Harj: Cool. How about if you talk about the Stripe story, the early days in particular a little bit. You were at MIT, then you left to start Stripe. How did you think about that decision? And obviously we’re in a stadium full of college students. How should they think about it? How do they know if it’s the right decision for them to leave college early and go start a company versus stay?
Patrick: Yeah. Well, I think I have the slightly unusual distinction of having dropped out of college twice to start a company. So maybe one thing to know is that it’s not totally trapdoor. You can drop out and in fact return. So I dropped out after my freshman semester to start the company with Harj. That was super fun. And then after a couple years of that, went back, did another year at MIT and then dropped out again to start Stripe. And when I went to college, probably like a lot of people here, I had this vision of my life involving becoming an academic. And I really like physics. And I thought I’ll do all this physics stuff. It’s so cool. I’d read all the Feynman books, all of this.
Well, growing up in Ireland, I hadn’t thought much about the possibility of startups. Hello to the other Irish folks here. Way back then in the pre-Cambrian era, startups were definitely much less well known even on campus and so forth. When I was dropping out, people thought it was super weird. I think overall, if you enjoy college, there’s no harm in finishing. I felt this real sense of urgency, which I think in hindsight was a bit unnecessary. But if you don’t enjoy college, it’s not your thing. It’s not what captivates you. You don’t really want to learn all the physics things or whatever. I think a lot of parents think that dropping out is very risky and will impugn your reputation for the rest of your life and so forth. And as far as I can tell, nobody has ever cared.
So I both think you don’t need to, but also the cost of doing so are de minimis.
Harj: What was the urgency you were feeling?
Patrick: The urgency?
Harj: Yeah, to go out and do something.
Patrick: I don’t know. Life is short. It was a general kind of haste. I think a lot of us, I’m sure many of the people here, get into this mode of speed running high school. And then once you get to college, it’s like, obviously I want to speed run that as well and do all the things. So it’s a bit of that. Marc Andreessen also talks about a version of this. I thought that a bunch of the opportunities in startups and in Silicon Valley and so forth were ephemeral and fleeting. And if we didn’t build it then, it wouldn’t be possible to do it in three or four years and maybe all the opportunities would be gone. In hindsight, I think that was a poor intuition. It’s been pretty robustly and reliably the case over many decades that Silicon Valley has a surface of opportunities.
Yeah, I think it was mainly those two things.
Harj: This is a very common thing that we hear when we talk to students now: part of the reason they want to drop out en masse, it seems at this point, is the worry that actually now is the moment. I think the meme going around is that if you don’t drop out and start a company and make lots of money, you’re going to be trapped in the permanent underclass. Should everyone here be worried about being stuck in the permanent underclass, I guess is the question.
Patrick: I think there’s a real... Humanity has always had an affinity for these millenarian models of how everything will soon come to an end and be this permanent transformation of society and so forth. Actually, there’s a great book, The Winged Gospel. People thought that after the invention of aviation, civilization was just entering—and humanity as a species was entering—a new era and nothing was going to be the same. Obviously, aviation was a pretty big deal, but I don’t think it was quite the sociological rewriting that some of the excitable proponents at the time imagined. So it’s hard to predict anything, especially of the future, but I would take the under on this being the last couple of years to create a company.
Harj: So going back to the Stripe story, Stripe ostensibly seems like a good idea. Even on day one, the internet’s a big deal, money’s a big deal, combine those two things. Presumably, is that how it went when you went and told people you wanted to start Stripe? Did everyone just say, “Hey, this is obviously a good idea.”
Patrick: It was kind of funny. Something we learned from YC was the importance of focusing on very concrete, easy to explain customer problems. It’s very easy to hallucinate or to imagine some customer problem that’s not actually something viscerally felt by a person who would pay money. And so over the course of, in part, working on Auctomatic together, we encountered this issue of it being really annoying to deal with the movement of money or payments or whatever on the internet. And on the one hand, it seemed like an obviously good idea in the sense that nobody liked the existing ways of doing so. They were broadly extremely unpopular and antiquated and legacy. You had to fill out this paperwork and go to the bank in person. The paperwork was in Latin and it was all bad. But then the flip side is, it just seemed kind of ridiculous that two kids would start a financial services business.
And FinTech didn’t exist as a sector at the time. The words literally didn’t exist. So we felt like the proverbial squirrels in a trench coat trying to masquerade as a real business or as serious adults, but obviously knowing nothing coming in about the space. Certainly, a lot of people we met and pitched—banks, partners, or whoever we talked to—they didn’t literally laugh us out of the room, but you could see them looking for the button to call security under the desk and have us hauled out because it just seemed so improbable. So anyway, I’d say it both seemed like an obviously good idea in that people really wanted this, but also a bad idea in that nobody took it seriously. But I think the fact that it was grounded in such a concrete, actual, real user problem saved us.
Harj: Actually, speaking of that, in order to actually build the product, you had to get a banking partner and do things that a typical software company did not have to do. As two young founders, how did you manage to convince a bank to trust you in the end?
Patrick: Yeah. Well, actually this is not an answer to your question, but just a thing that strikes me as I sit here is the reason we decided to start Stripe is because John and I were in college together. He was in his freshman year. We went to Startup School in 2009, which was held in Berkeley. We thought it was pretty cool. So we went and got sushi afterwards in Potrero. We were walking back from sushi and we’d been kicking around this idea for a payments thing or thinking about the space. It was walking back that evening after Startup School that we decided to start Stripe. That’s cool. I remember literally where we were in the road. I remember what we said to each other, which was, yeah, we might as well because it probably won’t be that hard.
Harj: Okay. So moral of the story is go get sushi in Potrero tonight and you might start the next Stripe.
Patrick: And yes, beware of these ultimate yak shaves. We thought we could do it on the side while in college. It would take a couple months. That was almost 17 years ago.
Harj: At the time, I remember you were also unusual in that you took longer to do a big public launch. Especially within the YC world, the motto is very much launch early, launch quickly, be out there and iterate. Could you maybe just talk us through a little bit about that? Why did you do it that way?
Patrick: Yeah. So we started working on Stripe seriously in the—well, we started working the week after that Startup School, but we were in college, it wasn’t full-time. We started working full-time the summer of 2010. We launched publicly September 2011. So almost two years after the first lines of code, after the repo was started. Waiting two years to launch seems—if we were going to YC meetings every week, I think we’d have been bludgeoned on the head. I think, look, in many domains, that probably is the wrong thing to do. In our domain, to your last question, because we had to do so much around security and partners and money movement and infrastructure and reliability and all the things, we didn’t feel like we could scale a really good self-serve experience without getting a lot of the preconditions and the infrastructure in place.
I think the thing that saved us and meant that it wasn’t a total walk in the wilderness is we had production users almost from the very beginning. So, first lines of code in fall of 2009. We got our first live production user in January of 2010, so two months into working or whatever. And it did very little. It was very larval and incomplete. Our first production customer was Ross Boucher at a company called 280 North. All it could do was charge a card. Ross would charge the card and then ask a reasonable question like, “How can I look at all my charges?” Reasonable request. So, let’s put up a little dashboard here. Then he’d say, “Well, I want to refund a payment.” All right, we’ll build refund support. After a couple of weeks, he was like, “At some point, do I get my money?” Also a reasonable request.
So, let’s build that functionality. It was very just-in-time development. Anyway, we had a production customer from very early. Then we did increase—it was in private beta—we increased the number of customers every single month all the way to that public launch. Every week we had actual customer feedback requests, new users coming in, we’re learning things from reality as opposed to our own hypothesized or extrapolated conception of it. I think if you have a significant stream like that of grounding, it’s probably okay to not be launch, launch. You’re an expert YC partner. Do you agree?
Harj: That’s a good question. Yeah. The issue with advice in general is it’s so generalized, and especially in startups, the exception proves the rule. If the cost of failure is high, then almost certainly you have to take longer to build. Maybe a slight tangent, but something I’m curious about related to this is we were talking about these coding agents, the ability to just build and produce software cheaply and quickly. I wonder, should people be taking more of this path? Should people be more ambitious in general with what the version one of the thing that they launch is? Or is it still fundamentally good product design to start narrow and focused and then expand out once you know what people want?
Patrick: Yeah. It’s a good question. I think probably in the era of AI—I don’t know. To some extent YC will be, I think, the expert here, but the whole traditional lean startup doctrine of exactly what you say: start out by buying the Google ads or something and identify this crevice or whatever and iteratively expand out from it. I think you can certainly imagine that that becomes much more competitive and much more aggressively tilled, and it’s hard to find those little niches. The internet’s a much bigger place than it was 20 years ago when some of those ideas emerged. Whereas taking these really divergent starting points where nobody else is trying to occupy that territory is maybe more—basically, maybe you have to more aggressively de-correlate in the era of AI. And I think it is interesting to think about many of the companies that are most successful over the last 10 years.
So many of them are very anti-lean startup, right? Whether it’s the labs themselves or Anduril or you can go down the list, a lot of them have this characteristic. So I think maybe a better way of saying it is 20 years ago, the whole lean startup thing was almost the only thing to do because of capital available and you didn’t have AI that made spinning up an organization with many different potentialities and capabilities so much easier. Whereas now I think you can start these much more aggressive and ambitious things upfront.
Harj: Within YC and probably the startup world at this point, you’re famous for at least the Paul Graham term, schlep blindness. Stripe, at least on the surface, involved a lot of schleps, things that presumably weren’t the intellectually most interesting things to work on. And I always found that especially interesting for you because you just mentioned you had academic interest in physics and clearly a deep intellectual curiosity and have very many things that you’re interested in. As Stripe has grown into this big company, in what ways are there intellectual rewards that you’ve given up and which ones have you gained?
Patrick: Yeah. In any company, there’s a bunch of stuff that’s not that rewarding or in and of itself all that interesting. Setting up payroll—no one starts a company so that you can set up payroll. And certainly building business financial services, there’s all sorts of more arcane and extensive versions of that. I actually feel extremely lucky with Stripe in this respect. And I think this is something, I don’t know if you need to think about it that much upfront, but I think once you think about it, maybe before you raise a significant amount of money, you always worry naturally about the possibility of failure and what’ll happen if you fail and how to mitigate and avoid failure and all those things. I think you need to ask the converse of that. What if you succeed? And you raise money and you have customers and you have employees and a whole thing.
Are you going to enjoy that? Are you going to want to work on that for 10 years, for 17 years, for 30 years? I mean, Larry Ellison at Oracle is going for, I guess it’ll be a half century soon, right? So what if you succeed? And in the case of Stripe, I really love it because we’re working with the world’s most interesting and innovative companies. 25% of all Delaware corporations are started with Stripe via Atlas. And then we get to partner with them and work with them and hear from them and get their feedback and get their requests and everything through the entirety of the journey up to being the Shopifys and the OpenAIs and all the standout successes.
Oh, and actually speaking of Atlas, we’re giving free Atlas incorporation to everybody at Startup School! If you are struck by the urge to found something over dinner this evening, as we were, just email startupschool@stripe.com and we will get you your link for free Atlas.
But yeah, I think PG latched onto something where there are all these menial tasks, but in the totality of Stripe, I find it so interesting. Every business is an applied theory on how some aspect of the world works or how some market works, or if it’s a new company with a new model, that’s a contrarian thesis on some counterfactual. I’ve never met a Stripe customer and thought that’s boring. So actually, the business as a whole has been the opposite of the schlep blindness instinct.
Harj: And you have a particularly unique perspective on this because you work with the big model providers of big lab companies and you work with all of the fast-growing AI startups on the ground. Something that came up a lot here yesterday, and honestly it comes up within the batches too, is people are just worried about whether their idea is going to get trampled by the big lab providers. Given your perspective, I’m just curious, how should people think about that?
Patrick: Yeah. Again, predictions are hard and certainly the labs are very competent, capable organizations. And maybe to separate a little bit, will rapidly improving AI capabilities do this or will the labs specifically themselves do this? I think in general, the track record of—no organization, if we go back 20 years, there was some of this sense with Google. When we were doing Auctomatic, the question was always for our company and every other company, what if Google does this? And Google seemed omnipotent and had this immense number of incredibly talented people and essentially infinite access to capital and servers and just all the things. Human organizations are complicated and it’s very hard to aggressively prosecute 100 different priorities and to deal with all the issues and interference that arises among them and so forth. Google has done incredibly well in a bunch of specific places, but it’s not like Google has done all the things, even if in some basic material sense, Google maybe had that ability.
So I’d say that the track record of that is checkered. And in general, I think that fear has been overstated. Now I think there is a more specific thing of just models themselves, forget the labs. Even if the labs aren’t particularly ambitious about expanding their scope. Just literally LLMs will obviate a bunch of, or agentic capabilities will obviate a bunch of specific verticals or tasks or something. Hard to say, obviously contingent on one’s forecast, it’s the model capabilities themselves. But in certain cases, I’m sure that will happen. And in certain domains it has already happened. Looking at the Stripe data, one thing I will say that I think is germane to people here, there are many more businesses getting started now than there were a year ago, as little as a year ago. Way, way more than were getting started five years ago. And actually the relative change between last year and this year is pretty much the largest relative change we’ve seen in any given year.
So for example, from 2019 to 2020, we saw a big jump, understandable during COVID. So February to April of 2020 or so. I think the growth rate inflected to maybe 50% or thereabouts year-over-year in terms of new businesses getting started. As I speak, the number of new businesses starting on Stripe is up around, it’s a bit under, but around 2X year-over-year, which again is the largest relative jump we’ve seen. And you might think, okay, fine. There’s way more vibe-coded, kind of lightweight slop, whatever. Fine, there’s more things, but are they actually succeeding? But actually, the median business is doing better this year than a year ago.
And then if we stratify it and look at the probability that any given business will reach some revenue threshold—a million dollars, $5 million, $10 million, whatever—those all seem to be getting better. The time-to-revenue for new companies incorporated with Atlas is declining. And so by all the objective metrics we can look at, it seems to be a better time than ever to start a business. And again, things can change. I don’t know what the world’s going to look like in five years, but speaking today on July 26th or whatever it is of ‘26, I think the Stripe data would suggest there’s never been a better time.
Harj: We see the exact same thing in the YC batches. Companies are just able to grow faster than ever. Certainly within the batch.
Patrick: Back in the old days when Harj and I were first starting out, getting to a million dollars of revenue—like runway revenue—was a big deal. People would know about that company. They’d be like, “I heard that X company got to a million dollars of revenue,” and now, I mean, that’s—
Harj: Yeah, you should be by your first month, it feels like. That’s an exaggeration for everyone here. But certainly within the YC part of the life cycle, like day zero to 90, it’s really being driven by, I would say, enterprises willing to buy from startups, which is the new thing. So you can sign these new contracts within the batch. You have the data as the companies keep growing. I’m curious, are there other factors that are driving these sort of inflected growth curves from one to 10 and 10 to 100?
Patrick: I think it’s really the dynamic you just mentioned, which is businesses everywhere are more spring-loaded to adapt and to try new things. And they have a real terror of being left behind with archaic and antiquated ways of operating. In normal times, you’re a new startup, you have some mechanism for doing whatever, and you pitch the CIO or the CTO or whoever at some company and they kinda don’t want to talk to you because your thing is not validated. Maybe you’ll be around in two years. All the obvious objections. But now people know that the risk of the status quo is actually extremely high. And so even if there’s risk in doing all the new things, this path also looks pretty dangerous. I really think there’s never been a better time for startups to sell and to have their products get adopted at pretty meaningful scale right out of the gate.
A lot of YC companies in recent times have demonstrated this, but I think it’s a really pervasive dynamic. And there’s a bit of it, I think, also. Stripe is not a consumer company obviously, but I think there’s some version of this on the consumer side where consumers are also pretty—consumers have complicated views on AI and maybe they don’t want the data centers, but people are very intrigued by the product. I think they’re kind of beguiled by them and there’s a predisposition and an openness to experimenting with the new.
Harj: Maybe just more broadly, something I’m curious about is, again, the data you have at Stripe. Has anything you’ve seen in that data changed a belief you have about AI broadly, say over the last 12 months?
Patrick: There’s a fear that AI is going to be this hegemonic, centralizing, totalizing force where a small number of companies gobble up a very large share of the economy. Many companies at the forefront of AI have done incredibly well and I think will continue to do incredibly well for sure, but based on what we can see at Stripe, the hunger and the intensity with which other companies are either getting started taking advantage of these new capabilities or existing companies are retooling, I don’t worry about the centralization in the same way. I think there are going to be many thousands of winners and, again, we try not to offer any definitive prognostications because the future is not predetermined, but based on the trend lines we can see, I think we are heading towards a more decentralized world and one with more broad-based prosperity.
Harj: Cool. All right. Well, I think that is all we have time for today. So thanks so much, Patrick, for being here.
Patrick: Thank you for having me. And it would be remiss of me not to say that Stripe would not exist without YC.
Harj: Oh, cool. All right. Thank you so much.