From Garage Nerds to a Million Deliveries: Inside Zipline's Impossible Bet

"If you're just getting started, you're an idiot." That's not exactly what you'd expect to hear from the CEO of the world's largest commercial autonomous delivery system.

But Keller Rinaudo Cliffton, co-founder of Zipline, isn't your typical tech executive. His company has quietly revolutionized healthcare delivery across four continents, serving 4,000 hospitals and health facilities with life-saving medical supplies delivered by autonomous aircraft. Just days before this interview, Zipline crossed a milestone that seemed impossible a decade ago: their millionth commercial autonomous delivery.

In this candid conversation with co-founder and CTO Keenan Wyrobek (creator of ROS, the Robot Operating System used by virtually every robotics company worldwide), they reveal the counterintuitive truths about building hardware companies, the near-death moments that almost killed Zipline, and why the most valuable companies of the next decade will come from nerds working on "weird" problems in their garages today.

Keller: My name's Keller. I started Zipline about 10 years ago. As a kid growing up in Phoenix, Arizona, I was a big nerd. I loved reading science fiction, particularly science fiction that was really hopeful about the future. As I grew up, I kind of started to realize that the future only gets better if we make it so. Like my grandparents were part of a generation that landed humans on the moon. They could also fly in supersonic aircraft. We don't have either of those things today. In some ways, technology moves backwards when you don't have really awesome talented teams of engineers working on a problem, striving to build a great product and make it available.

Keenan: My name is Keenan Wyrobek. I'm a co-founder and CTO here at Zipline. Anytime I've been able to make something where someone else just loved it and used it and it made their life better, I got that bug when I was very little. The thing I'm known for before Zipline is creating something called ROS, a robot operating system. It started as a PhD project to create an open source platform in robotics. ROS is everywhere. Pretty much every robotics company around the world uses ROS. So I spent 7 years on ROS, and that's when I knew I wanted to do something new.

It seemed obvious to us there was so much really impressive robotics technology in academic labs, but the thing that was more interesting was this idea of who's going to build the Apple of robotics. Why is this technology not making its way out of the research labs into our everyday lives?

Keenan: Robots really like boring problems. You want repetitive, you want controlled environments, you just do the same thing day after day after day. It's sort of like running water in that sense. It's not necessarily the most exciting thing, but you would definitely notice it if it wasn't working.

Keller: We felt there was just so much promise that if you could make logistics 10 times faster and half the cost, that would have a massive impact on humanity. We keep hearing about health campaigns that get stuck. The logistics causes it to get stuck or to fail. We visit a country and we're getting a tour of their warehousing. We see just a huge amount of boxes outside. Someone pulled us aside and said, 'Hey, that was all expired medicine.'

I think there were two key moments. One was meeting this researcher in Tanzania who had designed a system using cell phones. They had given cell phones to all these hospitals and health facilities, and they'd asked to text in when they had a patient who was having an emergency and they didn't have the medical product they needed to save that person's life. For pages and pages of the spreadsheet, he's scrolling. Outcome death, outcome death, outcome death, outcome death, and you look at the missing supplies, and these are all things they have in the country, all things we could deliver. And I remember thinking and asking him, 'Great, so what do you do when the text message comes in?' And he was like, 'No, it's just a research project.'

Keller: In those early days you had to be a little bit crazy. We all could have had respectable normal jobs. We didn't pay ourselves very much money for the first 5 years. Ryan was being recruited by a lot of great, famous tech companies that you've heard of, and he decided to instead move to Las Vegas and build robots in an apartment. Keenan had dropped out of the Stanford program in robotics. I think we all were sort of united by this sense of adventure, getting to build something totally new and crazy. To a certain degree, maybe psychologically, we liked that people thought it was impossible. We thought it would be fun to show that it could work.

The challenge is that for the first 2 years we got every door slammed in our face. Like nobody believed that this was possible. We ended up talking to the Ministry of Health in Rwanda. We're a team of 15 nerds. We're going to use these aircraft to deliver to every hospital and health facility in the country. We'll deliver all crucial medical products. And she looked at me like I was an alien and basically said, 'Keller, shut up, just do blood.'

And she told me this story that apparently they had had a mom lose her life in childbirth where they had needed a blood transfusion, and she went for 18 hours and they could not get a blood transfusion to her. And then we just happened to be lucky and walked in that afternoon talking about Zipline, talking about autonomous logistics. I think she connected it in her mind and said, 'Well, these nerds are probably crazy and there's probably not that good of a chance this works, but if this is a potential solution, maybe we should just try it.' And so it was like that one little opening that ended up leading to the very first commercial contract that Zipline ever signed, which was a contract to deliver blood to 21 different hospitals in Rwanda.

Keller: We were talking to potential healthcare partners based on pencil sketch of the concept. That's it. No technology, no drones, nothing. Most people think drone, they think something with rotors that hovers and things like this, we call Platform One. It looks like a large RC plane, but because we went deep with our customers first, we knew they cared about two things. They cared about range. They cared about cost. And if you're going to do those two things, it has to be an airplane. It cannot be a drone.

At the time, even engineers we hired said, 'Why are we doing a fixed wing aircraft? Why are we doing an airplane? Why aren't we doing a drone?' And we said, 'Well, because we know our customers. We know what they want, and this is the only way to get them what they want. If we make the other thing that everybody else is making in drones, we will fail for our customers.'

Keenan: I still remember the look on one doctor's face when I took a prototype to a health clinic and took it and threw it over the fence and it hit the ground and made such a loud whack. You could just tell from the doctor's face, they're like, nope. Not like that's not OK. And then, of course, we went from that to the box with a parachute on it that comes down slowly enough that everybody feels very comfortable with it.

From folks like me who come from robotics, we forget that most people are not like us. If people think, 'Oh, robot,' you failed. And especially in healthcare, our healthcare partners, they don't want us to deliver to their tech savvy patients. They want us to deliver it to all their patients. And so I always think of it as my grandmother test. If my grandmother would be OK with it, then we've solved the problem. To be clear, she won't, but she's not OK with most technology. So it's a very interesting engineering and design bar to achieve. But once you've got it and all of a sudden people like that are drawn to it and like it, you know, you're onto something.

Keller: I definitely remember this one moment we were still delivering to just one hospital in Rwanda, and I remember getting woken up at about 2 a.m., 'Keller, we have a problem.' It's not what you want to hear at 2 a.m. And so I said, 'What happened?' And she said, 'Well, we delivered this emergency blood transfusion into this inaccessible place on the roof,' and I thought, oh my God, I mean, clearly something had gone really wrong with the control algorithms. Like we had missed by 50 feet or more.

A couple hours later, 'Hey, I think we fixed the problem,' and I said, 'Oh, by the way, what happened to the blood? Is it still up there on the roof?' She's like, 'Oh no, no, no, the technician went and got a ladder and scrambled up onto the roof and got the blood, and they transfused it into a patient 5 minutes later.' And I remember hearing that thinking, wow, this is an example where we screwed up, but our customer met us halfway. It made me realize that if the problem is important enough, your customer will meet you halfway.

Keller: Today's Zipline is the largest commercial autonomous system on Earth. We just crossed our millionth commercial autonomous delivery a couple of days ago, but we actually have customers who want Zipline doing a million deliveries a day for them, especially in developed economies like the United States and Japan. Demand has just grown exponentially, and at this point we're signing new enterprise contracts with large customers, whether they be huge healthcare systems or huge quick commerce and retailers. We're focused on building the capacity that will be required from a manufacturing and a technology and an operations perspective that will enable that kind of scale.

I saw a flag in a gymnasium of a school a couple of months ago, and it says 'We do this not because it is easy, but because we thought that it would be easy.' It so perfectly describes the mindset of Zipline. We were like we can totally do that. That will be easy. And then once you are committed and you can't back out, then you actually have to realize all the ways that it's really hard and you have to solve those problems. That's the reason probably that a certain amount of naivete is required to start a startup like Zipline.

Keller: People often joke that hardware is hard. There's a reason for that. We have to be really good at electrical engineering. We have to be really good at firmware engineering. We have to be good at software, and that's just on the engineering side. Then we have to be really good at manufacturing, supply chain, customer success, and sales, really being good across all of these multidisciplinary areas. That's the thing that makes it so hard to build great hardware companies.

But the flip side of that is that when you look over the next 5 to 10 years, we're pretty confident that a majority of the most valuable companies that are going to be created are going to be hardware companies. I mean, you look at Apple or you look at Tesla or you look at Nvidia, you look at SpaceX, these companies have incredible competitive modes because it is hard.

Venture capitalists tend to be very risk averse, and they want to build ideas that are not capital intensive and that can just grow super fast and then sell or go public, and I think this leads to a lot of founders focusing on problems like scaling data infrastructure in the cloud or whatever. I think if we were to open our eyes and really ask ourselves the questions of what would it take to build the Star Trek version of the future that we'll be proud to hand to our kids, what are the important problems that we need to solve for 8 billion humans who live on Earth? There are a lot of these problems that require building hardware and building real products in the real world. I don't think people should be as scared of that as they maybe are. I think it's hard to build these kinds of companies, but these companies are also the ones that are going to move humanity forward in the most impactful ways.

Keller: There are a few key things. It's really important to fall in love with the problem rather than fall in love with the technology or a solution, because the reality is, if you're just getting started, you're an idiot. We didn't understand all the nuances and complexities of what was going to be required in a solution to really make this overall kind of company work. What was really important is that we went out and we were immediately trying to sell services and get customers to pay us money and we learned by doing, and we learned by having things go wrong under test circumstances. We crashed airplanes on our test site, we crashed a lot before we had a system.

Not being fancy, forcing yourself to engage with the real world, go and build a minimum viable product as fast as possible, see if you can get customers to pay you money for it. You know what nerds are working on during the weekends in their garages today are what will be the giant companies of 5 or 10 years from now. If you pick something that's obvious and that is well understood, then a company is just going to come in with a billion dollars of capital and dominate that market.

We needed to pick something that was far enough afield and weird enough that we could go become experts in that well before anybody else was paying any attention to that industry. I think you have to have that foresight. You have to go work on something that's weird and you have to be willing to be treated like you're a weirdo too, because people will be like, 'Why are you wasting your time on this? This is stupid. This is never gonna work.' You also have to have thick skin to be able to have people tell you that you're stupid for years and have the conviction that what you're working on is worthwhile. That's my advice.