Justin Kan, founder of Justin.tv and Twitch, describes the brutal early days: AWS costs were crushing them, almost no investors would fund the company except Tim Draper, and they had to build their own data centers just to survive. He recounts the 24/7 on-call stress and repeated outages that nearly killed the startup.
Chapters
I mean, there's like lots of— it's not so much one. Usually when you ask someone about like entrepreneurship, I don't think there's like one like time when it was like horror. I mean, there are oftentimes like these few horrifying times. I think I told some of the stories out there. But mostly it's like these small things that are like just horrible, you know, kind of like stressful. And then it's like the accumulation of them happening like every week. So, I mean, one thing that happened I remember that was, like, really terrible is, like, to save money— basically, we used to be— we were, like, a pretty big video streaming startup, right? We were, like, you know, one of the biggest live video sites out there as Justin.tv, and it was, like, extremely expensive, right? Live video is expensive to do. And so we used to— we would put it on, like, Amazon and EC2, AWS, which is what most startups use now. But then I remember one time in 2007, the AWS team called us and was like, you need to, like, turn down your bandwidth. 'Cause it's like too much for us. I think they've probably solved that problem by now, but at the time we were like, oh, okay. So we started figuring out what we could actually use, and we were too expensive, we couldn't use a CDN 'cause we didn't have the money. And when we went out to raise money, everyone except for Tim Draper and a few others were like, no, we don't wanna invest in video. Video's a bad investment. And so during this time, all these video startups died, like Revver, Veo, all these, You know, U.S. video startups, just like they went— they couldn't raise any more money and they actually went out of business. But we survived, and we largely survived by like trying to be like a cockroach and spending as little money as possible. And to do that, we ended up building our own data centers, which actually turned out to be this good competitive advantage in the end, but at the time was just like this tremendous amount of work. Like, I remember we moved from one data center to another and the whole team contributed. Went in vans that we rented to our data center, like, unplugged the server, took the site down, unplugged all the servers, put them in our vans, drove to this other place, assembled everything. And this was, like, now 12 hours in from, like— we started at, like, 11, and this was, like— or maybe now it was, like, 8 in the morning. And then we pressed the button, and, like, the site didn't work. Like, people were— like, my team was, like— I remember some of our employees were, like, passed out on the floor. They were, like, so tired, exhausted. And we were just like, what do we do? And, you know, spent the next 5 hours in the data center. I don't know if you've ever been in a data center, but it's super dry in there, right? It's loud. It sounds like fans. It's like a stressful environment on our computers trying to bring the site back up. And it was like times like this. Like there was another time I remember the power— like one of our breakers shorted in the data center and someone had to drive there. It was 5:00 a.m. on Sunday morning. Everyone was drunk. Like, there was no— like, I woke up to the emergency, like, someone calling me, and I was like, I don't— who's gonna go there? Like, I don't know. Like, eventually, I think my co-founder Kyle got up. He lived in, like, the closest, and he actually got up and went. I called him, but he was still drunk too. So it was, you know, it was, like, very stressful constantly. Mm-hmm. There was, like— those aren't, like, horrific incidences, but, like, it was, you know, our site would go down all All the time. People were on call like 24/7 for like 2 years. Right.
