A Driverless Car and a World-Changing Idea: Cruise Is On a Mission

‍In 2014, Kyle Vogt sold Twitch for over $1 billion. After spending eight years on a streaming platform for gamers, Vogt devised a stress test for his next venture: “If I’m not willing to work on it for the next 10 years of my life, I shouldn’t do it.” It wasn’t long before he and Dan Kan launched Cruise, entering the then-nascent race to build the first self-driving car. Kyle offers an inside look at how his business evolved from there, after Cruise pivoted from its original plan, got bought by GM (a deal that still allows Cruise to partner with automakers like Honda), and kicked off plans for a fleet of autonomous cars—if you can even call the Origin a car. 

“This is so obviously the future”

After Twitch, I decided that if I was going to do another startup, it had to meet a list of criteria I had devised for myself: First, I wanted it to have a direct and positive impact on people. Second, I wanted the business to succeed because of the technical problem being solved, not just because we have a good sales team. And third, I wanted scale. In order to solve for impacting people’s lives, you need as grand of a scale as possible.

I wasn’t actually thinking about robotics at first, but then I remembered the DARPA Grand Challenge for autonomous vehicles. The government put up a prize and had grad students scrambling to compete; it really bootstrapped the AV industry. I participated in it at MIT, and we did horribly, but it was a fun problem. If you think about the impact of self-driving cars, it hits everything: They’re electric, they prevent car accidents, they give freedom of mobility to people who are blind or otherwise impaired.

“It was the ideal manifestation of that lean startup approach: Find a way to use existing solutions, think about what the customer needs, get it done as quickly as possible.”

I didn’t want to go head-to-head with Google on a driverless car. So I used the lean startup mentality and asked, “What’s the minimum thing you can build, that would provide value to people, that they would pay for?” I homed in on this concept of taking an existing car and making it smarter. Put cameras and a radar on the roof, and a computer in the trunk. It’s essentially what Tesla Autopilot is today, but our plan was to retrofit any car with that capability. 

Within three months we had a car that could go down the 101 and move between lanes. It wasn’t perfect. There was no playbook for building a self-driving car at the time. Today there’s all these online courses, and off-the-shelf hardware components. I remember we were practicing for Y Combinator’s Demo Day, and we had 3D printed some parts that turned the steering wheel. We took it down to Mountain View and the car was sitting in the sun, and the interior cabin temperature probably got above 120 degrees. The parts literally started to go soft and melt. The victory moments were very fleeting in the beginning—quick futuristic feelings of, wow, I’m witnessing something that doesn’t exist, but this is so obviously the future, before we’d need to tweak something. It was the ideal manifestation of that lean startup approach: Find a way to use existing solutions, think about what the customer needs, get it done as quickly as possible to recruit more talent and raise more money.

“A self-driving car costs a lot, but fleets of vehicles that are totally driverless—that could work.”

Around this time, Uber and Lyft were blowing up, but had this terrible cost structure where they were basically losing money on every ride. We realized that the future of the industry can’t rest on building cars and selling them to people, like Google wanted to do. It’s about transportation as a service. A self-driving car costs a lot, but fleets of vehicles that are totally driverless—that could work. It became super clear that you can’t go all in on something like a retrofit business if the thing you really want to build is completely different, like fleets of cars. When that piece clicked, we were ready to abandon the retrofits.

Inside the halls of GM

GM has a venture capital arm with technology scouts, and they had a pulse on the automotive tech happening in the Valley. Originally they were just kind of sniffing around, but after they heard we jumped to Level 4 autonomy in just three months, they became very interested in Cruise. We ended up with a term sheet for an investment from GM. And we didn’t want to do it. It would’ve tied us to GM in a way that’s beneficial to them, because they get essentially sole access to technology. But there would’ve been a signaling risk for us. If you do a deal with GM, then other car companies are no longer interested in talking to you, because they assume that you’re tied to that one company. Eventually Dan [Ammann, former president of GM, now CEO of Cruise] came back and said, “Well, what if we buy you?” Because then we’d have access to its resources and the ability to integrate into its new electric vehicle. 

In hindsight, I didn’t spend enough time thinking about what Cruise would be like once inside of GM. The instinct seemed to be to treat us as more of an R&D arm. But ride-sharing is a nascent space, and the best practices are being figured out. You want centralized efficiency—a system where people can run quick experiments and fail. We ultimately made a compelling case for why GM should treat us instead like a brand new product, and they agreed to hand back this vertically-integrated responsibility to me, because I didn’t want to just build a technology; I wanted to build the whole business. It would have been hard for us to break through and fulfill that vision if we hadn‘t been given the freedom within GM to move fast, to make mistakes, and have that be okay—stuff that would typically make a large company squirm.

“I didn’t want to just build a technology; I wanted to build the whole business.”

In that light, the acquisition was a means to fulfill this vision of impacting as many people as possible. It was a way to accelerate our collective vision: GM wants to build autonomous cars and generate revenue from a transportation-as-a-service model, and so do we. With GM we could get there much sooner than if we were a standalone startup raising round after round of funding. This path gets us closer to the positive impact we want out of building this technology.

A fleet of cars, designed to last a million miles

The Origin is an awesome car, if you can even call it a car. After GM bought Cruise, we went into the GM design studio and they already had renderings for this kind of vehicle. I think people are converging around the idea that if you’re going to have shared transportation, it’s going to be a spacious cabin—more like a living room on wheels. As we got into thinking how this business was going to scale, we thought about how current vehicles aren’t well optimized for ride share. You’ve got your knees backed up against the seat in front of you. There’s a seat for a driver. The seats aren’t socially arranged. We knew that was suboptimal, but still, a better experience alone isn’t a reason to set off on a billion dollar journey. It wasn’t until we started getting really deep into spreadsheet modeling that we realized what matters in this business is unit economics. This goes back to the beginning: What was that Uber and Lyft problem that we were trying to solve?

As a thought experiment, if you handed Lyft some magic piece of technology that made it so they could hire drivers for one-fourth of the cost as Uber, then Uber would be dead. That’s what we’re doing with the Origin in plain sight—we’re building a car that has a fundamentally lower cost structure and a cool, futuristic experience. The Origin is very inexpensive, and it was designed to last a million miles, which is insane compared to a regular car. Everything is optimized to make the cost-per-mile of this vehicle a third or a fourth as much as any other possible way you could get around. We set off on that journey five years ago, and the first Origins will be on the road real soon.