1 Million Drone Deliveries a Day by 2028 | Keller Cliffton, Zipline

Today I'm sitting down with Keller Cliffton, the Co-Founder and CEO of Zipline. Zipline is a drone delivery startup, and I think the biggest drone delivery startup in the world.

Yep.

Yep. You guys have done a total of 2.8 million-ish deliveries to date, and then you just announced that you're aiming to do, like, a million a day in 2.5 years.

With one partner.

With one partner. And so I want to kind of start this off, like, I don't know if people realize how insane that number is, so I want to start us off with: what does this ramp actually look like over the next 1,000 days?

Well, you know, we're here in the factory where we're going to be producing all the aircraft that are required to make that happen. The internal perspective on how we make that happen, I mean, Zipline has been growing incredibly fast.

You know, we obviously operated outside the U.S. for the first 8 years of the company's history. Last year we finally got full regulatory—this kind of unprecedented regulatory approval from the FAA. We launched in the U.S. at scale, and the system just grew exponentially fast.

I mean, we grew 10 to 15 percent week over week most of last year and this year. That's growth in terms of flight volumes and in terms of revenue. As a result, we made an operating plan at the beginning of this year to scale the company to a million deliveries a day as fast as humanly possible.

And we built this factory in order to achieve that goal. The problem is that as we've been scaling across some of these metroes in the U.S., some big companies have been looking at what Zipline's doing and saying, "Hey, like, that would be very valuable infrastructure for our customers."

So case in point, you know, you mentioned Uber. I mean, Dara, you know, showed up here 4 weeks ago, looked around the factory, talked to us about, like, we ran some numbers on a calculator, and he was basically like, "I'll take all of it."

Because, you know, a million deliveries a day was the entire capacity that the company was planning for at that point when we signed that partnership. You know, as a result of that, I mean, Zipline needs to do a lot more than just serve Uber.

We have a lot of other amazing customers. Some of the biggest companies in the United States rely on us. And so it means that just 6 months after setting that plan, the plan is now being thrown in the trash, and we're now having to set a way higher target of the total number of deliveries that we need to achieve per day 2.5 or 3 years from now.

It has a lot of implications. It means that even though this, you know, factory that we're sitting in—I mean, you've gotten to walk around it—is, you know, it's big. We're going to need to go 5 or 10 times bigger than we were initially planning.

You know, the other implications are that when you see this technology growing in the real world, it actually sort of feels—just having now 2 years of observing how customers use automated instant delivery, we're realizing it's totally different customer behavior than we were previously expecting or you kind of saw.

How so?

Well, so, I mean, first of all, keep in mind, like, this is already a huge market,right? Like, you know, there are these very big companies, the big marketplaces, as well as, you know, more traditional retailers like Walmart or, you know, Amazon.

I mean, when you look at all instant delivery in the U.S., there are about 5.5 billion instant deliveries being done every year. And that's just instant, not including more traditional, like, two-day or parcel delivery. So that's a very large market, and it's not even including outside the U.S.

But the crazy thing is that if you were to just look at the customer buying behavior that we're observing in Dallas, in Houston, in Cleveland, in, you know, these cities where Zipline initially launched, and then extend that buying behavior to the rest of the United States, there would be 55 billion instant deliveries, not 5.

So it's like, you see customers who, instead of ordering a couple times a month, are generally ordering every single day. And it's just a result of having access to a system that is 10 times faster, extremely reliable, totally safe, magical customer experience, and less expensive on top of that.

You know, customers change their behavior. And actually, you know, interesting analogue to Uber in that I remember in the early days when Uber was launching, people would say, "Oh, well, even if Uber is only—it grows to be 50 percent of the taxi market in San Francisco, it's only going to be a $20 billion company."

And obviously, the thing they had wrong is today Uber is more than 10 times the size of the taxi market in San Francisco. So you make a new kind of service that's magical in a lot of ways. Customers totally change their behavior around it.

I was watching one of the interviews of Travis probably like 3 or 4 days ago, and he was talking about the user behavior of Uber. And he's like, "People, when they have a car available at their fingertips, they kind of change their behavior and how they interact with the city."

And so he was like, "We can't even predict exactly how much each user is going to use our product because they keep on using it more and more over time." And this almost feels like the exact same analogy again where someone uses the, like, instant delivery and then suddenly they use it, like, twice next week and three times.

Like, last week I bought multiple things on DoorDash in the, you know, in the same, like, 2-hour period in two separate deliveries. And I was like, "This is exactly the use case for Zipline."

Yeah, I think that'sright. I mean, Travis was very, you know, farsighted in that prediction. You look at, like, what happened with Uber, I mean, he was exactlyright. Like, all of these new—I mean, people now commute in Ubers that you never would have done that with, like, the traditional taxi industry.

And so that's obviously, like, a giant market expansion that was kind of unlocked. And in this exact same way, I mean, Zipline, you know, when you—when I'm spending time with customers in these different markets, I mean, you know, you meet a dad who has, like, 5 kids and he's ordering from Zipline a couple times a day.

They'll order multiple things, like, even over the course of an afternoon or evening. You know, 80-year-old grandma who's ordered from Zipline 350 times in the last year. I, you know, just posted on X yesterday this funny video I saw of, like, it's basically a rooftop, like, an apartment rooftop frat party that's happening in a pool.

Oh, I saw this.

It's like a seven—yeah, it's like the seventh-floor rooftop pool deck of, like, all these kids just hanging out listening to loud music in their swimsuits, and they're ordering Zipline, and Zipline is delivering directly next to the pool on the seventh floor of the apartment building.

And so you kind of realize, like, whoa, these are just totally new kinds of, like, different ways of engaging with delivery.

Something that I also loved about that video was there's, like, very few products where everyone just kind of is astounded by, "Oh, this thing actually exists."

Yeah, what the hell am I looking at?

Yeah, on that video, every single person had stopped what they were doing, whether they were in the pool or, like, sitting on the side, to just look up at the drone delivering some package. And my bet is that has, like, a very strong emotional reaction, and suddenly they want to go, you know, use the service themselves.

Yeah, I mean, a lot of the—you know, one of the reasons we've been seeing this crazy growth, I mean, you know, a lot of these different—you know, we're sort of launching, like, neighborhood by neighborhood in these cities, and a lot of these neighborhoods, I mean, you know, take a guess, like, there's a—I mean, some of these cities, we have more than 50 percent of all homes ordering from Zipline.

There's just, like, insane levels of customer adoption in just 6 to 8 months. And you kind of ask, like, "How is that happening?" And the reality is it's a very visual product. People love, like, people are using it when they're at soccer practice.

And, like, one of the moms will be like, "I'm going to order, you know, whatever, a Gatorade and oranges and then have it deliveredright here." And, like, the entire—like, you know, you have 100 people, like, "What am I seeing?"

And, you know, and then they can download the app and use it themselves and get their first delivery that night at home. You see, you know, birthday parties, like, it's, you know, it's a hit. You're delivering to people's backyards.

I think that, yeah, the product is inherently viral, I guess, is the short version.

I love this line, "Everyone has a plan until they get punched in the face." And I feel like if you look back 18 months at the beginning of last year, I don't think it's almost possible to predict, you know, 15 to 20 percent week over week growth for an entire year or an entire 18-month period.

And so I got to imagine that the plan 18 months ago was completely different than it was, like, 12 months ago, 6 months ago. And today, you know, one customer walks in and they take your entire supply that you were planning to build for maybe multiple years.

How has that plan kind of shifted in the past, like, 2 years? And how are you kind of planning going forward to basically predict for unpredictable demand?

I think that, you know, the reality—people say hardware companies are hard. The reality is that the challenge with hardware companies is you have—I mean, you have 30 different disciplines,right? You have everything from electrical engineering, the teams that are building the actual flight computers and the avionics and the power distribution board and the battery management system.

You have the firmware teams who are writing the low-level software that runs on those boards. You have the powertrain team, which is designing the electric motors and the batteries and then all of the, you know, busbars in the aircraft to actually transfer energy throughout the aircraft.

And then you have the mechanical engineering team, which is designing the 43 major subassemblies on the aircraft. You have aerodynamics, aeroacoustics, guidance navigation controls. You have all the different software teams owning everything from communications architecture to multivehicle deconfliction to, you know, detect and avoid, like, computer vision-based detect and avoid.

You have the customer ordering application. And that's just engineering. Then you also have to be really good at operations. You have to be really good at logistics. You have to be really good at manufacturing. You have to be really good at supply chain.

And so you can just imagine all these things as, like, a really long chain. And you are as good as the weakest link in the chain. So you can imagine you can be really good at 29 of the 30 things.

You're like, "Oh, man, we nailed it on 29 things." Manufacturing, we just didn't nail it. It's like, cool, the company's a zero. You don't have vehicles. Any one of those things is a zero, the company is a zero.

They're all multiplied by each other. And I think this is the reason that maybe it's, like, only the paranoids survive or, you know, if you weren't—if you didn't have, like, psychological problems before starting the hardware company, you definitely will afterward, you know?

Because I think it breeds, like, a very healthy degree of skepticism, paranoia, and humility. It is really hard. And especially once those kind of companies enter hyperscale, you're now in a mode where you have, like, hundreds of millions or billions of dollars of commitments in a supply chain.

And those are, like, hard and fast financial commitments. You have a factory that you have bought or are renting at, you know, usually factories aren't cheap. The factory has to operate at extreme scale in order to make sense from a unit economics perspective.

You know, I think it's all hard. And then, yeah, as you mentioned, you also are sort of engaging with the real world, which is you're sort of suddenly realizing, "Oops, our plan was off by 5x. Our plan was off by 10x.

We actually have to do 5 times more than this."

Or, like, your timeline is completely off. Like, you thought that you had 5 years to scale and you actually only have, like, 500 days or.

Yeah.

Yeah.

I think all of those things are true. And so I think it's the reason, you know, people talk about, like, hard filters and soft filters and the Fermi Paradox. I don't know if you, you know, talked to anybody about that, but, like, yeah, I think you could think of it as, like, the filters.

The Fermi Paradox driver.

The Fermi Paradox driver companies. Where are all the hardware companies? Like, why did they all die? You know, oh, they died because there was a—like, there are multiple hard filters. And getting these companies to scale is hard and capital-intensive and also requires a lot of luck, I think.

But the reality is when you then look at a lot of the most valuable companies on Earth or the companies that are playing the biggest role in moving humanity forward, you know, a lot of these companies have giant hardware components, if not are primarily hardware companies.

And so I think that it's not for the faint of heart, but it's also important for humanity.

What is the thing that has caused you the most stress in the last 12 months?

It's a different thing every month. I think, you know, to my exact point, like, there's no—you know, you're, like, combating this insane fire that will for sure kill the company if it cannot be fixed quickly. And then, you know, you're attacking that thing with total focused, ruthless, you know, energy, and then 2 weeks later it's onto something completely different.

That is no longer the bottleneck. The bottleneck is here. Now we got to go focus on that. So we needed to launch Platform 2. Then there was a huge amount of just, like, integration work that we had to do to kind of, like, shake the rocks out of the gears, actually get the system operating reliably last year.

Then the system started to scale exponentially fast in terms of flight volumes. We had to figure out aircraft utilization and a lot of—there's a huge amount of improvement we had to do to kind of, like, the customer-facing app and how we integrated with all these different partners throughout a metro.

Then, you know, tariffs were announced and this caused, like, you know, some pretty significant disruptions in the supply chain. Then we were having a problem manufacturing enough aircraft. Then you get all the aircraft into the field and you realize, like, oh my God, we got to maintain all these aircraft.

And so Zipline has had, like, multiple existential maintenance crises over the last year. Fleet health today is the best it ever has been. A month ago it was the worst it had ever been. So it's just giving you a sense for, like, you know, it feels like you're living in this heightened state of—it's just, like, incredible sensitivity to timelines and having to, like, attack these problems one at a time.

So, and what will be the existential crisis for Zipline a month from now? Nobody knows.

Interesting.

If I knew we would be focusing on it today.

Brian Chesky has this interesting idea of you basically, as the leader of a company, have to be scaling faster than your company is, or else you're kind of, like, falling behind and you're kind of becoming the limiting factor.

For something like this where the growth is just getting pulled forward like crazy and the expectations are rising insanely fast, how do you kind of stay ahead mentally?

You know, the reality is I think it usually feels like you're sort of, like, holding on for dear life, you know, by your fingernails. I think that the way I've thought about it for myself is that I'm learning as fast as I possibly can.

I'm being humble and curious in any given year. And usually I'm totally failing and having to figure things out, you know? And maybe by the end of the year I feel like, oh wow, I finally feel like you're Neo, you're seeing the ones and the zeros in the matrix, you see how it all works, you're like, I can.

But then the game changes in some sense.

And then, yeah, and then a week later the company doubles in scale and you're instantly back to being useless and unprepared.

And I think the key thing is being humble enough to listen to the people around you who can tell you when you're being stupid, who can point out, like, all the things that you're doing wrong. You know, I—and then also I think it requires, like, a certain degree of shamelessness, you know?

Like, so much of Zipline has felt like I'm saying, oh, let's just, you know, this is the answer, this is the direction we're going in, everybody follow me. And then a month later it's like, whoops, that is not a good idea at all.

Everybody back. Like, let's go in the complete opposite direction. You know, this isn't working. We need to do something totally different. And I think you can, like, beat yourself up for bad decisions. You can look backwards. You can be like, oh my God, I look like an idiot.

Why would people even trust me? You know, I think it's better to not look backwards. I think just, you know, practical problem solving, below ego, and try to figure out on any given day, like, what are the things that are not working and then how do we—how do we fix it as fast as possible?

I think when you kind of go through this hypergrowth phase, the company has to radically change how it operates and how it's, like, structured generally. Like, you may need to increase your manufacturing capacity by, like, 10x over a year.

And that's a completely new function that you've never really done before. So how is the company changing over the next, like, 12 to 24 months?

I think that, you know, there are certain things that Zipline focuses on that are not changing. Like, the culture of the company, focusing on practical problem solving, focusing on hiring people who are hungry, low ego, practical problem solvers, mission-driven, fast learners.

Like, we never fail when we hire people like that. In general, we've tried to set the company up to be small special forces teams that are moving with high degrees of freedom and responsibility. We believe in having as little, like, middle management.

I mean, basically, like, no middle management. I just think, especially in a world where velocity determines the success of the company more than anything else, and in a world with AI tooling and in a world with, frankly, just better communication tools like email and Slack and other solutions, the idea of having middle managers who need to, like, pass information up and down the chain is useless.

And so we prefer to have small teams led by entrepreneurial leaders who have a huge amount of responsibility and operate, like, you know, with CEO-level responsibility with regard to the thing that they are owning. That isn't really changing.

I think that, you know, as the—as projects become more and more complex, you do have to, like—you don't get to be a shit show startup with, like, all these teams running in totally different directions. And so I think, you know, you have to try to get really disciplined in terms of how do you, like, set clear goals and then think about those goals across these teams.

So, like, assuming that they all get—you know, they all add up to one unified coherent vision that results in a complicated product launching into the real world, it's also particularly critical for things like safety and reliability. Like, those are cross-program requirements that we have to—on every day that we're going and serving customers, we are attesting to our regulator and, more importantly, like, to the communities that we serve that this system is going to be safe.

And it means that Zipline's had to build a huge amount of infrastructure. And I think this is one of the ways that the company really had to level up in the last few years. We built a huge amount of infrastructure.

Some of it is actually here in this building. Huge amount of testing that we do, component-level testing, aircraft-level testing. You know, we can—we have chambers here where we can make it rain, where we can make it snow, where we, you know, simulate intense heat and humidity.

And then we go from there to our three large test sites. So we operate in the real world 24/7, 365. These test sites are a huge part of the way that we can move very fast from a hardware and software development perspective while then validating the overall system to an ever-increasing safety bar.

You know, like, it used to be enough to be able to say, oh, yeah, like, something's only going to go wrong one in 10,000 flights. And then you got to get to one in 100,000 flights. Then you got to get to one in a million, one in 10 million.

You're looking for increasingly rare problems. And it requires increasingly clever or extreme ways of testing systems to their limits so that we know how it's going to perform when we put it into the real world.

When you first came back to the U.S., like, from Rwanda, and you started doing deliveries here, I imagine that that product experience was completely different than the thing that exists today. And you also don't necessarily want to scale something up a bunch if it's, like, not super great.

What was kind of the flip in your mind for this is, like, good enough that we do, you know, it's like a 9 out of 10 or 9.5 out of 10 customer experience and this is what we want to scale?

I, you know, my first reaction to your question is actually hardware companies generally don't really have that luxury. I think software companies do have that luxury. You're, like, trying—you know, you deploy an app, you're like, oh, it's not good.

Like, let's learn from it. Hardware companies, honestly, I mean, at least for the first, you know, 3 or 4 major platforms, which will probably take the first 10 or 15 years of the company's life, every new product is a complete existential bet for the company.

If it doesn't work, the company is certainly dead. You know, it's not like if Tesla on deploying the Model S, it turned out the Model S wasn't a very good car. I mean, Tesla doesn't survive. The company's dead.

If the Model 3 winds up not being a very good car, even though Model S had been a profitable program, Tesla's probably dead at that point because of, you know, the crazy amount of debt that the company had taken on building the gigafactory and, you know, scaling up that production line.

So there's this obvious sort of goal of, like, you find your beachhead, the use case where you can find customers who will actually put up with a surprisingly bad customer experience. I mean, that is kind of a thing,right?

If you look at the Roadster, and the Roadster was, like, a Lotus Elise chassis and then they put, like, a battery designed by a different company into the trunk of the Roadster and it's like.

I think there was, like, hundreds of separate suppliers and.

Yeah. Oh, the car, I mean, the car was incredibly unreliable. They lost a lot of money on each car that they sold. And yet, you know, they were able to show that there are customers out there who are willing to buy an electric car, which I think at that time was a totally counterintuitive idea.

Yeah. And in the same way, you know, Zipline focusing on blood transfusions for the first 4 or 5 years of the company's history, we had found this teeny little narrow beachhead where we knew the service wasn't going to be perfect.

We knew we were going to have to get these unprecedented regulatory approvals. We knew we needed a customer who would put up with just a lot of, like, iteration and imperfection. But we had, you know, it was kind of clear.

It's like, oh, but every flight is saving someone's life. So the customer will probably kind of partner with us on this. And we wereright about that. And so I think, and then you go from there, you know, you can look at the, you know, parallels for either company, but for Zipline, going from that to, like, getting Platform 1 to the point where today it serves 5,000 hospitals and health facilities and saves 17,000 lives a year.

And then, you know, we had a lot of big companies in the U.S., like Walmart and Chipotle and Cleveland Clinic and many others saying we would use that. Like, we think this would be fundamental infrastructure in the United States that we could build our businesses on top of.

That gave us a fair amount of conviction. Like, there's something here. You know, if you could automate delivery from any building directly to any home, that would be a really big important system and layer of infrastructure. But when we were launching Platform 2, I mean, if it didn't succeed, the company would have died for sure.

And you have no idea if you're going to have product-market fit.

What have been the strangest challenges that you've run into?

I mean, there are a lot of weird things that happen in the real world, you know? Like, we worry about nesting, you know, which you can imagine, like, that as an engineering term or a software development term. No, we're talking about, like, critters, like bugs and little animals, like nesting in infrastructure that you build in the real world.

You know, we worry about, like, actual bugs and raccoons and birds who will be, like, nesting in some of this hardware. You know, you worry about the level of crazy weather that you encounter if you're operating across many different countries 24/7, 365.

And you've kind of told customers, like, we're always going to be available to you. It's not very compelling to say, like, oh, yeah, Zipline is this life-saving service. You can totally depend on us with your life and the lives of your loved ones when it's sunny out.

Not very compelling. Like, you basically have to fly all the time that people are going to depend on you in that way. And so, you know, I have a lot of healthy, much healthier respect for, like, what does it mean to fly in a hailstorm?

What does it mean to fly in a lightning storm? What does it mean to fly in insane dust and rain and, you know, up to 40 degrees Celsius and down to negative 20 degrees Celsius? Not only that, you don't only have to worry about terrestrial weather.

When you're building these kinds of systems, you also have to worry about solar weather. So that's a good example of something weird because it turns out that weather on the surface of the sun causes really intense electromagnetic interference with things like GPS systems on Earth.

And so you have a lot of this radiation that can basically interfere with the signal that's being passed from a GPS satellite to a GPS unit on an aircraft. And we rely on something called dual-band RTK differential GPS.

It's basically like military-grade GPS that only became commercial a few years ago.

Do you have to basically use, like, a GPS denied, like, flight controller or something?

Well, that's a whole other thing that we also worry about, which is GPS denial. This is, like, this is more like GPS interference.

The satellite link gets knocked out or something?

Yeah. Well, it's just solar weather happens in cycles. And it happens over the last year. We've been in the worst part of the cycle. And it has a huge impact on, like, how you operate these kinds of systems and how you harden the systems to things like that kind of solar weather and solar radiation.

And so I think the list goes on and on. Like, we could talk more about all the weird things that humans do in the real world, like weird human behavior that you have to plan for, both, like, genuine mistakes as well as, like, people sort of trying to, like, hack the system and, like.

How people tried to hack the system?

I think just really simple things like, you know, all of the weird things that people might do to try to, like, defeat safety systems or try to get, like, a video of the system doing something wrong. The great news, actually, for Zipline is that because we are primarily delivering to, like, people's backyards, it's actually a lot easier for us from a safety perspective than it is if you imagine, like, scooter companies where you're just, like, on a sidewalk generally annoying people and there's very little accountability.

Like, you know, people can grab the scooter and, like, throw it into the ocean or something. For us, like, we know who we're delivering to. We know, you know, we're delivering to someone's backyard. If something, it is, A, very difficult to, like, interfere with the system because obviously the main aircraft is staying 100 meters in the air when it's delivering to you.

And, B, people don't have very much incentive because, like, the thing is creating this magical experience for them, a lot of value for them. And it's like if someone were to interfere with the system, it's like, okay, we're showing up on your doorstep, like, 5 minutes later asking you, like, what happened?

And so the lack of anonymity also helps.

You know, with Soren and Olaf at Neros, their product is actually a drone and, like, delivering a payload, whereas here the product is actually just a thing appearing, like, kind of out of thin air in a very short period of time.

Does that enable you to make, like, special iterations on the actual hardware in the sense that you don't necessarily have to, like, sell a customer on it? You can just make whatever decisions you want and then the thing just appears faster?

Yeah, I think it's a really good observation. I mean, when you look at a lot of hardware companies, you look at a lot of particularly robotics companies, very easy trap to fall into, which is, like, you're sort of in an ivory tower or you're kind of, like, building in an R&D lab and you're building, like, a super cool robot or you're building a really, really cool vehicle.

I think the reality is, at least what Zipline discovered very early on by trying to launch in the real world was that the vehicle was 15% of the complexity of the solution that was required. Like, when we originally launched, we were immediately getting punched in the face, as you said, by reality.

And, you know, the punch in the face is realizing, like, oh, wow, we need, like, all of these layers of software integration. We need a plan for maintenance. We need a plan for inventory management on site. We need a way of providing an unmanned traffic management system to the regulators so they know where these aircraft are at any given time.

We need to build a customer ordering portal. Like, all of these things had to not just be designed and then deployed, but also made reliable because at the end of the day, none of Zipline's customers want to buy a drone.

I think this is, this is, like, the fundamental realization, which was that there are so many industries that are going to be revolutionized by robotics, but I think that it is unlikely that companies are going to be, like, selling robots to these companies who are then going to figure out how to manage the deployment and handle the integration and manage the training and manage the maintenance, etc., etc., etc.

It's just, like, none of these companies want to be operating a giant fleet of autonomous aircraft. So instead, you know, we are selling teleportation. We are selling instant delivery that is 10 times as fast and less expensive and dramatically better for neighborhoods and a better customer experience.

And all our customers care about is does something go from point A to point B fast enough to save somebody's life or create, like, this new kind of commercial transaction that wasn't possible before and really delight customers. And they want to think of us very much the way that they would think about working with any other logistics company.

You know, you pay money and the thing is transported from point A to point B.

Ideally, you don't even realize that it was a drone. It just kind of, like, plops out of there.

That'sright. They don't care. I mean, if we actually had, like, teleportation technology, would they even really know the difference? Like, they wouldn't, you know? Like, we, yeah, I think it's kind of a powerful point. And then to answer the second point of your question, it does mean that Zipline, by keeping a lot of that complexity sort of behind the curtain, on one hand, I think sometimes it's probably true that Zipline's customers may not even fully appreciate the level of, like, engineering and manufacturing scale and regulatory challenges that are required to make this kind of a system work.

It's one of the reasons, you know, one of the big questions we get from investors is they'll often ask, like, well, some of the biggest companies in the world have been investing, like, a billion dollars a year in drone delivery and they have very little to show for it.

So, like, why is that? And I think it's just that the problem is much harder than people think of. On first pass, they're like, oh, I can go buy a quadcopter and duct tape a Snickers bar to the bottom of it and fly it a mile and I've just done drone delivery.

And it's like, okay, but big difference from that to operating a new kind of infrastructure layer that can do millions or tens of millions of deliveries a day.

It's like the best version of Amazon is basically you just click a button on an app. You don't know anything about the infrastructure behind the scenes and the thing just appears, you know, the next morning.

Exactly. Yeah. I think that that is how Amazon feels to people today. And obviously, there's this huge movement from all of these companies into sub-30 minute and then sub-10 minute delivery. I think logistics is going to feel more and more like the internet where it's just like, yeah, you click around, press a button, and you can instantly get the information you need.

I think that in the same way that the internet can deliver information, you know, robotics will deliver everything else.

One of your core values is just customer obsession. And if you want to have a really great customer experience for every single person that's on the network, you may have to basically stop people from signing up as, you know, until the hardware can kind of, like, scale.

So how have you guys kind of had to, like, stall growth in order to keep the customer experience the level you want it to be?

Yeah. I mean, someone must have given you a hint on that question because actually that is one of the weird things that we are doingright now. Yeah.

Oh, okay.

So we are having to, you know, we built the system so that we could serve, like, a certain range and a certain number of homes, assuming a certain level of, like, utilization per home. And then we ended up being off by an order of magnitude in terms of the utilization.

And so we've actually had to constrain the number of homes that we're serving, sort of artificially constrain because we don't have the capacity. I mean, the systemright now, you know, our charging sites are open. The system is operating from 6:00 a.m.

to 11:00 p.m. We will soon be 24/7 in a lot of these markets that we're operating. And most of our mature markets, like, the system is max utilization for almost every hour of the day. And so we are now scrambling to deploy more capacity and try to catch up with that demand.

But it is really hard because, you know, it's sort of like we definitely do have our version. I don't remember if you do you remember the Twitter fail whale? It definitely feels a little like that. I mean, if you're using the Zipline app in some of these cities, we are often you're encountering a message that says, like, demand is really high.

We're struggling to meet that. You know, we're struggling to meet it. We're trying to add more capacity. Please try again in 15 minutes. I mean, that sucks. That's hard.

Well, this is what the, like, the Cybercab feels like. Every single time I open the Tesla RoboCab app, like, taxi app.

They just don't have capacity.

It's just not available. And I'm like, okay, the 15th time you do that, you're kind of like, I'm just going to open Uber.

Yeah, exactly. So, I mean, we worry about that. And I mean, the crazy thing is Zipline is deploying giant amounts of capacity into the real world as fast as we possibly can. We're running this factory at ever-increasing scale and we still can't keep up.

I mean, I think the reality is, you know, on the demand side, you know, when Dara was here a couple days ago, he mentioned, he's like, look, the thing that Uber has always known about demand is that, especially demand for instant delivery, is that it is insatiable.

I thought, yeah, I mean, it's true. It's a good word. And the reality is it sort of feels like, you know, you've accidentally put a straw into the Pacific Ocean. Sort of have this sense of, like, whoa, this is going to take, like, we got to think bigger and this is going to take a long time to actually satisfy that scale, especially given that instant delivery is a very big market today and it's going to grow by another 10X in order to serve this.

And by the way, on the 10X piece, I think the thing that should be obvious and also exciting is that, I mean, we are already hiring millions of humans to drive 3,000-pound gas combustion vehicles to deliver something to you that weighs, like, an average of 3 to 5 pounds.

And that is obviously very expensive. It's slow. Sometimes it's unsafe. Sometimes there are reliability challenges.

Road congestion.

Yeah. You cause traffic. You cause pollution for neighborhoods. Like, all of these things. I mean, the reality is actually you don't even necessarily replace that. It's just that I don't think it's going to be possible to scale that by another 10X.

I think that that's a really important intuition here, which is that, like, there are not nearly enough humans or cars or, like, capacity on streets in our neighborhoods and cities, which is a limited, which is truly, like, a hard and fast constraint.

There's we cannot scale 10X the way we're currently doing things. The only way to do it is through, you know, robotics, physical AI, and using new kinds of infrastructure.

When I was talking with Denton off camera earlier, he was mentioning that you guys could increase the, like, area that a given, you know, drone or Zip,right?

Yeah.

Yeah, that a given Zip can, you know, access or maneuver in, but it's not actually necessarily beneficial because it's already being utilized, like, 100% in that small area. And so is there kind of, like, any trade-offs that you're making on deciding to not necessarily, like, improve the platform to just scale faster and, like, get more out there because you know that even if it's a small zone, it'll just automatically get 100% utilized?

Yeah. I mean that, you know, it's just, again, you're attacking the bottleneck. Right now the bottleneck is not demand. So we're not rushing to add more. Like, you know, we get so many customers who are like, why, you know, why is my home not enabled yet?

I mean, Zipline just, like, this week enabled a record new number of homes. So every week we're adding more and more homes. We're basically taking homes off the waitlist and adding them to the service. Butright now the challenge is capacity because customers are ordering way more than we were expecting.

And so we're just, we're having to deploy more and more capacity in order to be able to serve the homes. What we don't want to do is be in a mode where, like, 70% of the time you come to the app, we're like, demand is too high.

We can't serve you. So we're really trying to, like, yeah, you're just trying to catch up. And soright now the constraint is capacity. And that's the reason that we are, I mean, even while you and I are sitting here, like, this is the main aircraft line that we're sitting on.

This week it is getting disassembled and moved into the next manufacturing bay over because we are doubling the size again. So we're just moving as fast as it can be possible. I mean, if you come back in 30 days, this is all completely, you wouldn't recognize the building.

So, yeah, just moving as fast as we possibly can to try to attack the constraint of the month.

How many factories are you going to have to build out to actually do a million deliveries?

Just depends how big the factory is. I think that, I mean, this building, when you get, you know, when you get to walk around it is pretty serious scale, at least serious scale for the Bay Area. You know, this building was originally designed to be able to produce something like 20 to 40,000 aircraft a year, which was our initial estimate of what would be required to satisfy, like, a million deliveries a day.

It's now clear that we need to build something like we need to, we really need, like, five times that capacity or more. And so Zipline is now starting to figure out what does it mean? Where are we going to build that manufacturing capacity?

You know, it's all happening in the US, but we have some big investments to make.

If you fast forward, like, a year or two, is there any major changes that you're making to the platform at that point that you won't make today, but you kind of already know is, like, here's something that we want to add or change?

I mean, the way I would probably answer that question is, you know, when we were talking with our team, we were meeting with a lot of the different engineering leaders on the team, and I actually showed them this thing that I had seen Jeff Bezos present in a presentation he gave a couple years ago, which was he actually showed the, a screenshot of what Amazon.com looked like in 1994.

So it's basically like in the first couple months of their operation.

It's got a little river.

It says, like, Amazon.com. Yeah, I think it has a river. And then it's basically like these blue hyperlinks. It's just, like, straight up, like, a vertical list of blue hyperlinks with the different kinds of books, like sci-fi, romance, fantasy, nonfiction.

And then you click it and then it's just a list of all the books. And that was, and that was the website. And so I feel like it would be a bit like asking, you know, Jeff in 1994, you're like, oh, wow.

Like, can you imagine, like, what are the changes that you could imagine making to this website? You're like, you know, I do think, and, you know, they always emphasize, like, it's 12:01 a.m. I mean, when it comes to autonomous logistics, you know, this is, I think we are going to, in the same way that I'm sure people at Amazon today look back at that screenshot, they're like, wow, can you imagine how ghetto it was?

Can you imagine, like, what humble beginnings? Like, I think that we look at everything that we're doingright now. I mean, there are a thousand things that we want to make better about the system, you know, like you're, and you're every day by operating in the real world, you're seeing all the different ways.

You're like, okay, we could double the efficiency here. We could dramatically change the way we do, you know, completely change the maintenance model. You can imagine different ways of assigning capacity dynamically amongst these sites. You can imagine totally different kinds of ground infrastructure that would, like, triple the density of capacity that we can install in any given charging site.

You can imagine more efficient ways of deploying charging sites. You can imagine different kinds of sites that we would be building. You know,right now we're mainly kind of focusing on renting little bits of parking lots and building charging infrastructure that's sort of similar to Tesla supercharger.

We're just building different kinds of, you know, we're building chargers for aircraft rather than for cars. The closer you are to the technical implementation, the more embarrassing you find it. Like, people, people look at it from outside and they're like, wow, like, that is so impressive.

That looks so sci-fi. And to us, we're like, you know, I don't know. The closer you are, the more obvious it is all the, you know, the 50 different ways, the a thousand different ways that we would want to make it better.

And then it's just prioritizing your list and you're like, okay, well, we can focus on, you know, 15 or 20 things in the next quarter. So, like, let's do these 15 or 20 things. So every, I mean, the, I guess the simple answer is just like, there's a long way to go.

Like, these kinds of hardware platforms, I mean, imagine, you know, the iPhone 1 and then all the way to what is the number we're at today? I actually don't know. You know, every year there are, like, extraordinary gains.

And I think these gains compound in ways that humans sometimes have a hard time imagining. We're just not that good at imagining exponential curves.

I watched a video probably like a couple years ago on, like, parking lots and how many parking lots there are around the United States. And it's kind of funny to me that at the same time that autonomous vehicles are taking off, you also have the instant logistics taking off.

And it seems like maybe this is kind of like this magical serendipity where you can just suddenly take a bunch of the parking lot space, convert it into instant delivery locations or something.

I think that that's totally a thing. And to me, you know, and I also this tells you something about, like, I think the technologies that are going to win versus not win. And you can also see this with, with data centers.

I think there are all kinds of interesting parallels. But in my opinion, you know, we got to be heading toward a future where we are giving the Earth back to humans. And so there are lots of examples. I think maybe scooters fell into that category.

I think in the way we do instant delivery today kind of falls into that category where you're causing, like, more and more traffic on roads just to deliver something to you. I think that.

It's like unhappiness of the human whole.

Yeah. I think sidewalk robots fall into this category of it's like, okay, well, we've already caused, like, max congestion on the roads. Now let's, like, also take over the sidewalks. Like, the last little public space that was meant to be reserved for humans.

Let's also, like, take over that with robots and force the humans, like, jump into the bushes. I think that we got to be, you know, technologies have to be scaling in a way that makes neighborhoods, like, more beautiful, quieter, safer, less pollution, less traffic.

I think that putting data centers into space, like, is an interesting example. I think that there are lots of different ways that robotics and autonomy can kind of contribute to that. Zipline has always really been of the opinion that, like, yeah, you want to build a kind of delivery that gives the Earth back to humans, give fundamental, you know, whether it's parking lots or roads or sidewalks, like, all of that is for humans.

You know, we can use, you know, we can, we can transit in ways that are hard to see, very difficult to hear, totally unobtrusive. And you're just sort of moving logistics or, like, other commercial applications into a different plane.

You can make cities and neighborhoods way more magical and way more human-centric by doing so.

AI data centers have basically, like, the worst favorability ratings out of anything in the world. And I think when we talked last, you said that even things like DoorDash and Uber Eats have, like, a negative score, but somehow are growing anyway.

And you guys have, like, some crazy 85, 90, something like that on, on NPS. How does that change the dynamic of scaling in new cities and geographies? Like, where the AI data centers, a governor might go, I'm not going to let people build here.

And suddenly they get elected because people don't want them. Whereas this is something that people do definitely want and everyone that interacts with it kind of has this, like, magical experience. And they're like, I want that, you know, for meright now.

Why doesn't it already exist?

I don't really understand it. I think that, you know, we need to be, we need to be actually building a future that people are excited to hand to their kids. And we need to have clear, optimistic visions of, like, sci-fi, sci-fi futures that we want to build.

I think that, you know, I watched Eric Schmidt give his, like, commencement address in Arizona, you know, and I thought it was so crazy to just observe, like, he clearly went in thinking this is going to be awesome.

It's going to be a banger.

Yeah, it's going to be a banger. Like, exactly. He's like, they are going to be so excited.

You are a tool for this.

They're going to be so excited to hear what I have to say. And, like, these kids are going to be so excited about the role they're going to play in the AI future. And I think it just, you know, it's interesting to just observe, you know, I think there are people who are really smart, who are maybe significantly miscalculating, like, the temperature in the room.

And I think the temperature in the room has to do with, like, a lack of clear vision on the people's parts who are building the technology to say, like, these are the values, this is what we're doing. So I think it's really complicated.

I think at the end of the day, like, we need to be building versions of the future that are, like, awesome for families, awesome for normal people, helping people achieve the goals that they want to achieve in their life.

To me, you know, I do think, I mean, I'm a, I'm a techno optimist. Like, I believe very strongly that the most likely outcome here is that we are going to build abundance for humanity. We can have a hundred billion people living on Earth.

And I think the Earth is going to feel like a nature preserve.