Zach Rash & Daniel Singer, CEO & CBO of Coco Robotics, on why ground delivery beats drones

With delivery platforms paying $50B annually to drivers amid rising labor costs, companies across the industry are exploring autonomous solutions ranging from aerial drones to ground-based robots.

To learn more, we chatted with Zach Rash & Daniel Singer, CEO & CBO of Uber and DoorDash’s ground robot partner Coco Robotics ($36M Series A, Sam Altman & Founders Fund).

Key points from our conversation via Sacra AI:

Coco was founded in 2020 specifically to tackle last-mile delivery. What was that original moment—the problem you were really fired up to solve that inspired you to start Coco?

Zach: There are a few things. I grew up in Menlo Park, where DoorDash started, so I saw the gig economy and food delivery start pretty early. I had this thesis from when I was in high school, and then in college when I worked on robots. It felt like the future we wanted to build was probably not your neighbor in a 4,000 pound gas car delivering food millions of times a day up and down the street. That felt economically understandable—you're tapping into underutilized labor and supply, giving flexible earning opportunities. Those cars are already there, people already have them. So I get how we got here, but it didn't seem like the end state we want for our cities.

I always thought it was inevitable that we'd have a more purpose-built solution for movement of goods that's not someone's car. It can be much lower cost than a car, but more purpose-built for the task and fully automated. That seemed like a natural evolution as AI progressed. You're going to have a purpose-built vehicle for cities autonomously moving anything and everything around those cities. That's going to benefit the merchants and the local economy significantly, the customers in the form of more capabilities—they can do more transactions that way. And ultimately, that was going to have a huge impact on the city because you can take back more space that was used for parking.

You can get more cars off the road, which has significant emissions reductions and noise pollution reductions. You're generally going to beautify the city if you build it more around people than cars. And you're also going to get this really cool future where, if you make the delivery of goods extremely low cost, the city will start changing. You'll start storing goods closer to customers because that last-mile cost is so low. Real estate starts becoming either optimized around fulfillment and delivery or experiential.

Today, we're in this hybrid between those two where every restaurant is a fulfillment center and supposed to be a hospitality experience. So that middle ground is really ugly and not very functional and not a perfect experience for anyone. Our vehicle and technology can be a huge part of making this just way better for all those constituents. So that seemed inevitable to me, and I felt compelled to start a company in this space even though there was a handful at the time. I felt the approach others were taking wasn't the right one, and there was an opportunity to do it a lot better.

Over the last five years, what has been the biggest technology or market change that has most impacted Coco Robotics since you started?

Zach: AI for sure. It's been crazy to watch. I was doing AI research in school before we started this, mostly on reinforcement learning and how it would be applied to real-world robots. We were obviously betting on some sort of AI progress and trend line—that autonomous vehicles were going to be a reality, and we'd be able to navigate complicated city environments autonomously. That was always a key bet that would happen.

But the rate of progress there has just been incredible. One of the key things we did was, if you want to bet on that rate of progress, you need to be well positioned to take advantage of that. You need a lot of data, and you need to have a real-world service where you're very good at managing a fleet out in these real-world complex environments.

I never really liked the approach of building a super expensive car with no customers and doing test runs on test tracks. That scales very poorly. It's super expensive to develop things that way. But if you get this crazy progress in AI, you need a massive fleet to be able to take advantage of that because you need a lot of data and a lot of human feedback on that data. So getting a real service to scale with real customers has huge benefits as AI keeps progressing.

That trend line has been awesome. All the intelligence you've seen getting developed in the software world and now moving into this agent world—that same thing starts translating into the real world in a very big way because whoever owns and operates these machines that are capable of this level of intelligence are just going to keep getting smarter and more capable and able to do more things at a lower cost.

I would love for you to differentiate why Coco Robotics' approach is better. First, your approach versus an aerial approach in terms of the mode of delivery. And second, you mentioned there are a few different models in terms of robotics delivery—why is Coco's approach differentiated or better?

Zach: Let me clarify that the term "sidewalk robot" is used often. I'm not a fan of the term because it focuses too much on a sidewalk and the fact that the sidewalk is where you operate. There's a reason we need to be able to use the sidewalk, but the way to think about the form factor is really a bike courier. This is the autonomous vehicle equivalent of a bike courier. And a bike courier is the best way to move goods around a city.

That's why almost every city runs delivery by bike couriers. Asia, Europe—it's going to be an e-bike. It could be a moped. But most of those markets run on bike couriers. Most of the densest markets in the US run on bike couriers. I'm sitting in Los Angeles, which has no bike couriers, not because it wouldn't make sense here. It's because you have to have a car to live in Los Angeles, so everyone's using what they already have.

We can use roads, shoulders of roads, sidewalks—depends on the local legislation in different areas. The vehicle can go up to about 15 miles an hour. So we're using a bunch of different modes to get around the city, but the important part is to be useful to a merchant in these super dense urban settings, you need to be able to get very close to the merchant, if not inside the store, and then the same thing for the customer. If you say you're going to find the next place where you can double park or stop with an autonomous vehicle in Manhattan or Downtown Chicago or Brickell in Miami, the restaurants will not use this. Tony Xu from DoorDash commented on this in the last earnings call. He got a question about how Coco has been working. And that was one of the big points he made: I don't think autonomous cars are the right solution for delivery of goods. I actually think you need a new type of autonomous vehicle, and that's obviously what we're working on with them.

From a traffic perspective, a parking perspective, a merchant experience perspective, it's really important to think about how do we design this to be more like a bike courier and less like a car.

Daniel: But as compared to aerial stuff, I have a unique perspective coming from the aviation world. I am a pilot in my spare time, and I have a deep fascination with aviation. The way those overlap with both aviation and the delivery business—the constraints of aviation tend to inversely map to the problem of moving things around, which if you start thinking about food, this is less clear. But as it becomes the general business of moving things, it becomes super obvious.

When you want to move things around, generally you want to move more of it—more weight, more volume, and generally more value. Especially on the weight, useful load, and volume question, those are categorically opposed to what all the constraints of aviation allow you to do.

On the second side, a lot of these delivery use cases really depend on reliability. Everyone thinks about the obvious one, which is the labor savings and cheaper vehicle. But almost more important to anyone who does delivery or logistics is that 98th, 99th percentile is where you lose all the money. That is where you make and lose all the money. So being able to be 98, 99, close to 100% reliable makes a huge difference to your customers and partners.

The thing intrinsic to aviation, and I'll apply this specifically to drones and delivery, is even in commercial aviation today—big heavy aircraft like the 737—to do it safely, you're still not anywhere close to those reliability rates. You get delayed. You get canceled. You get rerouted. From an infrastructure standpoint, we do not have the infrastructure in this country around air traffic control to really deal with this problem. And this is not for lack of financial resources. All the airlines and the entire commercial aviation business is constrained on air traffic control for the most part, and a little bit on the hardware.

Where this really boils down to is it's a nontrivial regulatory hurdle, and it's really hard to imagine in New York a bunch of drones going around. I'm in Jackson, Wyoming—I could see a drone being useful here. But the costs are nontrivial of operating aviation. Everyone tends to think, "Oh, they're drones. They're cheap. They're simple." Even with electric hardware, when you have things in aviation, the redundancy and safety and all these overheads you have to add to operate aviation safely, it just becomes much, much harder versus when you look at all the big logistics businesses, it's how do I have a really simple platform? UPS is a bunch of trucks with a bunch of people driving them. It's relatively simple from that constraint, and you're making that work.

The same thing is we're able to utilize a bunch of this existing infrastructure far more efficiently using this purpose-built ground form factor at a much higher reliability rate. Rain, sleet, snow, high wind—doesn't make a difference. And you're able to scale that platform across distances. It might look a little different depending on the distance. I don't think it won't work, but if the whole constraint of the industry is you need to get to lower cost and higher reliability, it's hard to see anything better suited than the way you've been approaching it.

Zach: Two really quick things to add to that: the on-demand movement of goods and trying to make that more affordable is going to be a really tough value-to-weight problem. Drones are really good at high value-to-weight. Zipline doing deliveries of medical emergency equipment in Africa is a genius business. It makes a ton of sense there, and it's very cool. Areas like Nebraska or more rural areas where you have to cover really long distances, it can start to make a lot more sense if you can solve a lot of the problems Daniel talked about, which are significant hurdles.

But in a city, it's super inefficient to try to move goods that are heavy. If you're trying to move a couple pizza boxes and a two-liter bottle of soda, you're going to be burning through a lot of power because you have to levitate that thing, drop it down to the customer, and then go back, which means you need charging infrastructure to handle that. So you're either going to need to way over-deploy capex to have enough vehicles to meet the throughput of your merchants, or you're going to just be doing a very small number of them and constantly be recharging, both of which are bad.

From the product experience perspective, being able to drop it down in someone's yard in a rural area is awesome. That's a magical experience. Doing that in a city—Daniel's got a big landing pad over there to drop his package. Doing that in a city, I don't know where you drop it. My food will get stolen in a second if it's left on the curb unprotected. And then what are you going to do, wait for the customer? Well, it takes on average about three minutes for the customer to get the order once we arrive. So you're going to hover for three minutes? You already have to charge after every single delivery for a drone. So the longer you have to hover, and our 90th percentile wait times could be 10 or 15 minutes, you have to hover that whole time or you're going to leave it unattended.

Daniel: From a merchant standpoint, I have to do some infrastructure build-out to be able to handle this.

Zach: You might get a Walmart or someone who is willing to invest in that sort of thing. But the capex, given the battery constraints, are really significant.

Can you paint a picture of how Coco integrates with existing infrastructure? When you say it integrates with existing infrastructure, with aerial drones you have to have these huge docking areas in the Walmart parking lot. What existing infrastructure is it that Coco Robotics is integrating with?

Zach: In the last year, even the last few months, we went from tens of merchants to many thousands of merchants across LA, Chicago, Miami, and Helsinki. And we didn't do any training. The robot just shows up like a driver. The merchant gets notified. They come out. They put the food in the robot, and it travels using some combination of bike lanes, shoulders of the road, sidewalks, crosswalks, and gets to the customer.

That's pretty magical. We just dropped autonomous vehicles into these cities, and then it just works with thousands of new merchants. These are largely restaurants—a very hard business, very busy, high turnover of staff. They're trying to take care of their guests dining in and take orders and manage the front of house and go out and do this with the robots. Successfully doing that at that scale that quickly was pretty amazing. So it's very hands-off from an onboarding perspective.

We spend a lot of time on our product side and technology side making very high-resolution real-time maps of the city. This isn't a map where we're taking a LiDAR scan. It's stuff like: Is this sidewalk segment pedestrian-dense by time of day, by day of week? Is there a car blocking the sidewalk? Is there a traffic light out? Is there a big snowbank? Is it flooded? Is there construction? We have a real-time map of all of this, and we route accordingly so that we can get from A to B in the fastest possible way for a robot.

That map gets updated in real time by the fleet. So we always have the most efficient way to operate in that market. Some markets are easier than others. Navigating Helsinki is delightful. Navigating Venice Beach, where I am right now, is a little bit harder.

In the future, you start using roads and bike lanes a lot more and maybe the city of the future has more dedicated lanes for this, but we're certainly not waiting for that to happen. But it's inevitable as you get this to more scale that there starts becoming a little bit more dedicated infrastructure.

One of the first areas we launched was City of Santa Monica. They did this pilot called the Zero Emission Delivery Zone where they took downtown Santa Monica and tried to say, "Okay, no logistics-related cars." So they wanted delivery robots, cargo e-bikes for Amazon packages and stuff like this. It was a huge success. The product just naturally uses way less energy. It's 10 times more energy efficient than a Tesla Model 3. So it is very low energy to move these goods around, which is really important.

But they were looking at how can we repurpose some of the infrastructure for a world where we want to get rid of cars, or have fewer cars related to logistics. So that was a first step in that direction.

You spoke earlier about how the cost is really in getting from 95% to 99%. If you could break that down for me and explain why that last bit in terms of getting close to 100% is where the big cost reduction has to come in.

Daniel: Can we use an illustrative example? This doesn't necessarily correlate to real numbers, but just to illustrate the point: What is the delta between 98% reliability and 100% reliability? If I make a dollar of contribution profit on every order and I do 100 orders, and my average basket size is $50 fully loaded, and two of those orders at 98% have some sort of on-time and full issue where I have to refund the customer or it doesn't get there, the food's cold, whatever—I'm not making a profit anymore at 98%. And at 99%, I'm doing $50. The changes there—it's massively different. You can move those numbers around, but to illustrate just how much that long tail and the 90th, 99th percentile makes a huge difference in terms of your delivery partner and your customer because those refunds when you mess up really blow out the profitability on the rest of your orders. So being able to be really reliable really does matter.

Zach: It's a huge impact on economics, but also, to benchmark this against UPS or Amazon or something, your window to be on time and full is really tight because you're moving hot food. You're moving prepared food. So you have a 15-minute window or a 20-minute window. Having five nines of reliability on a 20-minute window in the most complicated urban settings on the planet is a really hard problem, and that is a battle of thousands of hardware and software and operational problems every single day that you just need to be excellent at.

A lot of this business is combining software, hardware, and AI technology with human operations to make this an extremely operationally excellent business. And doing that well at scale in this category is super hard.

With DoorDash or Uber, you're relying on the incentives of entrepreneurs to go figure everything out. You have the merchant and you have the driver, and they can just problem-solve things together, and you get some sort of completion rate and success rates. The benefit of autonomous vehicles is you can massively improve that because you control everything end to end. But the challenge is you have to be really good because you're responsible for every single decision and every second of that trip. You don't have just two smart humans trying to figure it out.

So the ceiling of possibility is much higher, but it's a very hard engineering problem to get there.

What's a target here? One in 1,000 failure rate? What would be a sweet spot business model-wise and tech engineering-wise?

Zach: For something like this, you're just going to always want it to get better and better. But you certainly want to be approaching at least three nines. I don't have the exact numbers on this, but I would imagine human driver failure rates are somewhere between 95% and 98%. Some markets with maybe more severe weather are probably much worse. It's harder to get drivers on the road, so you have a lot of limited supply, so you're going to constantly miss your estimates for people. So that's probably where the industry is for food, something like that. Don't quote me on that number because I don't know the exact number.

But anecdotally, from my own experience and from what we see, that's probably about right. So we definitely want to be above two nines, but you're trying to approach three and then more from there. Obviously, getting five nines is tough because some of those things are outside of your control, just operating in a city environment.

Is this market big enough on its own with verticals that you're already strong in, like restaurants? Or is this dependent on many or quite a few adjacent verticals being able to adopt a solution like this?

Zach: We're trying to build a general purpose platform to move anything and everything autonomously. We started with food, hot food delivery, because it is the most difficult and most expensive thing to deliver today, so it's where our product can add the most value. Super short time windows to make the delivery, which makes it really expensive. So it felt like the area that the product can make a huge impact. We wanted to start by tackling the hardest problem. The cost comparison on packages is much lower than dropping off a hot food order.

So we started there. Now is that a big enough market in itself? It is surprisingly large given how expensive it is. This always blows my mind. Food delivery is on average at about a 60% premium to going to the restaurant to get food delivered. That's enormous. It's already expensive to go to a restaurant, so you're talking about a 60% premium on top of that.

So at that price point, DoorDash alone is currently paying drivers $16 billion a year. That's driver earnings on DoorDash a year. Globally, they're probably 20% to 30% of the market. Maybe non-China, you're looking at 35% of the market. So you're probably looking at somewhere around $50 billion or so of driver payouts just for food delivery at the current price point.

So if you make it 10 times cheaper to get food delivered to you on demand, that market will grow dramatically.

So that's enormous. Now we're not addressing all of it. We're not doing deliveries across rural Nebraska or something like that. But in the US, I would estimate roughly half of that's urban. In Europe, it's probably 80% of that's urban. Even the suburbs in Europe are incredibly dense. So it is a vast majority of those orders are in markets where we have designed a custom form factor to be the perfect delivery modality for those markets.

We're also focused—those urban markets are also 100% of the current problem with delivery. It is where all the congestion is. It is where all the extremely high cost of living is, and so it's where there's actual unit economics challenges. It's where all the regulation is happening. So the urban markets are also the most acute—it's a majority of the market, but it is also vast majority of the actual economics challenges and where the most headwinds are for delivery.

You can see in the California cities, you have Prop 22 increasing driver earnings. Seattle has something similar. New York's continued to increase the minimums for drivers, and then a lot of the cities in Europe—Germany has reclassified gig workers as full-time employees, so that cost has gone way up. Spain did the same thing. This constrains supply while increasing costs.

So those city environments are going to get much worse than where they are today from a cost and quality of service perspective.

So that's an enormous market. Now we are only just starting there because it's the most broken. We already today do groceries. We're starting to do more retail and packages. We will be moving anything and everything, and that expands it into the trillion-dollar general logistics industry. But starting with food and grocery, you're still looking at a pretty large business.

Daniel: When you start imagining further out, the brilliance of Coco, and this expands to a big philosophical thing that you'll see is different than the rest of the market, is there's this religious focus on providing value today and something that actually solves a problem today rather than hanging out hoping this will be useful in the future. But if we look at that future, where are all those other use cases?

There's roughly 4.5 million storefronts in the US when you include food, retail, etc. There's a huge demand for moving things around, and especially as you drive that marginal cost much closer to zero. Even intra-store—I was just giving an example in the food space. The amount of times if you have multiple locations that you're sending things from one store to another. Currently maybe it's not worth it to take an employee who you're paying hourly to go take them out of there and move that thing. There's so many other examples in shopping, etc., that as you bring that cost down, you can empower all these business owners to do that.

That's a really interesting imagination experiment that I always do: if you could, would people use it? And the answer is increasingly yes. But that's the beauty of starting. The market's already big as it is. We could do nothing else and be an amazing business. But as you accrue these advantages—bigger fleet, cheaper cost to operate, etc.—it becomes really exciting what you can do across verticals.

In terms of servicing these things and maintenance, I understand that's another differentiator for Coco. Can we talk about that? I think of the scooter boom that we had. It seems like at the end of the day, that was the big hole where a lot of the cost went and a lot of the reasons that didn't work out as well economically was because of the maintenance and charging costs and all that stuff. So if you could just speak to that less sexy part, what happens after delivery?

Zach: This was one of the things we spent a while thinking about before we started the company because Brad and I, my co-founder and I, were charging Bird scooters in our apartment in college. So Bird started in Santa Monica and Westwood, and we started doing this at UCLA. Scooters evaporated after three weeks at the most. So that would be a huge problem for us if that was to happen. A robot's more expensive than a scooter. So that was one of the first experiments: are we going to have the same issue?

On the actual vandalism or more accelerated depreciation side of things, we have seen almost nothing. That's over millions of miles of traveling in very busy areas.

The reason is we're not leaving them out. They're stored at night at the merchants or in local storage pods. They're parked at merchant property all day. Then they go to the customer and they come back, and they have some sort of remote supervision during that time period as well. And there's cameras all around it and a speaker and a microphone. So deterring crime is much easier than for a scooter.

The default mode of operation is it has a chain of custody the whole time. Somebody's always watching it and who has an incentive to protect the fleet. So just it doesn't really happen. Also, in most of these neighborhoods, if you go and try to damage a scooter, you're just hurting some invisible tech company. If you go damage one of these things, it's got your neighbor's family's food in it, and it's delivering from the local business down the street. So maybe in some neighborhoods that matters less than others, but generally, you don't get the kids and people kicking them over for fun.

So most of the maintenance cost is just on the day-to-day operations of the fleet: charging, cleaning, swapping out components that wear down. So one of the reasons and one of our competitive advantages and differentiations, we have by far the lowest cost vehicle in the market, both in terms of production cost and total cost of ownership.

Not only is that expensive to actually produce, it is then very expensive to maintain. Your depreciation per hour is going to be really high, but then your maintenance cost is proportional to the bill of materials. If you have a high bill of materials, you're going to have a high maintenance cost. And because the vehicle is smaller and more lightweight, it is easier to manage the fleet, swap the batteries, do repairs, do the pickup versus a full-size car.

So our fleet management costs are fully profitable today, and that part of the business, we're very good at.

I'm curious about how you manage the cost around human supervision and then what the path to autonomy is and if that's the one path to super profitability.

Zach: There are two major costs in this business. One is the fleet management cost, which is the cost of what we just talked about, upkeeping the fleet, but also the depreciation on the hardware. That basically needs to be affordable, and then you need to get really good utilization on that hardware. So when you deploy it, it needs to be doing trips all day every day. And to touch on micro-mobility for a second again, the utilization on these things is over an order of magnitude higher than micro-mobility because they can reposition themselves. They're accepting orders at a certain rate across different merchants and so the utilization is way higher, and then their revenue per order also is probably higher. So that's one side of business.

The other side of this is how much labor cost you have—what's your variable cost? How much variable cost you have per trip? And that's mostly remote teleoperators. Because you're always going to have some. You don't really want a world where there's no one that can do anything if the fleet runs into challenges. This is a totally under-discussed part of autonomous vehicles. Every autonomous vehicle company today has somebody watching. I would suspect most of the autonomous vehicle companies today have one person watching at all times, even if the vehicle is driving itself autonomously.

For us, what we designed the system to do is say, "Yeah, I want to be able to have one operator manage multiple at a time." And as the reliability of the autonomy software gets better and better, I can improve that ratio. We got very good at how we manage the teleoperators. So we use software to switch them between orders, between countries. Anytime there's any downtime, that teleoperator can go get a different set of robots. So we're very efficient with the labor and the time. We really optimized that side of the business when we started the company.

And then now as you have more of a world of autonomous navigation, it's extremely efficient with that labor. So as the AI solves harder and harder edge cases, you will be able to say, "Okay. I'm expecting an intervention much less frequently. So I can give them 10 robots at a time and not expect any material delay to any of those deliveries."

It's not just about can the autonomy get this from A to B. It's going to get to A to B at least as fast as a teleoperator could, and can it do that 99.9% of the time? And if you're watching 10 robots at a time and one stalled out and you add five minutes of delay to the other 10 orders, that's not usable. You can't use that product. You will be refunding orders, and you will be getting crushed. So it has to work incredibly reliably. And this is in the areas—this is not a college campus. This is not a suburb. These are in the most dynamic parts of the world.

So combining the human operator with that AI is one of our core competencies, and it's one of the most important things to be successful here. It's just not visible to most people, but that's such an underappreciated part of the business.

In terms of competitors, there's Serve, there's Starship. I'm just curious about what's unique to Coco. Is this operational efficiency and lower cost robots that we've already talked about? Is there anything else that you think differentiates your approach?

Zach: In general, we took a focus at the very beginning of the company of being operationally excellent, and that doesn't just mean people operations. That's how do you combine software, hardware, AI, and all of the people process to run a really high-quality service. We wanted to be good at that. We wanted to be good at the operations and the service part first, and then you add autonomy over time. And a lot of other companies in this space did the opposite.

If you see some companies that focus mostly on less dense areas or campuses or indoor environments and stuff, it's because they said what technology works to be autonomous, and then what environments can we actually deploy that to even if those environments aren't very interesting from a business perspective or aren't solving a huge problem.

Food delivery on the campus is only interesting because it's being subsidized by the school and by people's parents. You're shuttling food 100 feet from a dining hall. There's huge value in bringing food from the neighborhood onto into the dorms and onto campus. We do that today. We do a ton of deliveries on the USC and UCLA and some of the universities in Chicago and Miami, but we're not building a campus product. We just have access to the campus.

So I'm not opposed to going onto campuses, but I never liked that go-to-market. But it was because they wanted to focus on where can a 2017 autonomous vehicle stack work decently well? And it's not that hard to make an autonomous vehicle shuttle food 100 feet. So that works, and they found a way to make the business model work.

So this translates into all of our decisions and how our product operates today. And the result that you can see—this all sounds nice, but the result you can see is we have the largest urban fleet, to my knowledge, delivering for both Uber and DoorDash.

We are one of the first companies to be trusted to actually scale up, and it shows in the numbers. We're also able to profitably run these deliveries. So we can make money on our orders, which delivery in general was historically hard to make money on. But autonomous vehicle delivery is really far away from making money in most situations. So we work really hard on being disciplined to actually want this to be a viable business today, not 10 years in the future.

In terms of your partners, help me get my head around it. Are you almost always working with a DoorDash and an Uber as an intermediary, or are you going door to door to these local businesses and signing them up for Coco? How is that coming together?

Zach: We have a combination. We do partner with the platforms where we're basically a network of drivers on their platform so we can get matched to their customers and their orders that flow through. That obviously is great because it gives you enormous scale.

But a number of merchants do want to control the fleet. They want to brand it. They want to customize it. They have enough demand that they can actually just manage this fleet, and it's way more economical to buy robots from us and to operate the robots than it is to staff a full-time driver.

So we have the opportunity to make it much more affordable for them, but you can think of it as we're actually making it affordable for these restaurants to have in-house drivers. So it's a massive improvement in quality of experience and brand and all these things they care a lot about at an affordable price point. And that price point is less than a human driver. But that product is really interesting because it's not just a cost equation. That stuff's really important to these brands.

Are you doing both today?

Zach: Yeah, to be clear.

Are you able to say about how large your fleet is just so people can have an idea of the scale?

Zach: It's around 1,000 vehicles. We do have new production going right now, so we'll be well over 1,000 at the end of this year. We'll have around 10,000 next year.

On unit economics, are you able to be unit economic profitable maybe in some locations or in some cities or in some situations? You're not saying that you're unit economic profitable on a per delivery basis on average yet?

Zach: We are, globally today. Now, we plan to continue investing in growth, but the fundamental unit economics are working.

This transcript is for information purposes only and does not constitute advice of any type or trade recommendation and should not form the basis of any investment decision. Sacra accepts no liability for the transcript or for any errors, omissions or inaccuracies in respect of it. The views of the experts expressed in the transcript are those of the experts and they are not endorsed by, nor do they represent the opinion of Sacra. Sacra reserves all copyright, intellectual property rights in the transcript. Any modification, copying, displaying, distributing, transmitting, publishing, licensing, creating derivative works from, or selling any transcript is strictly prohibited.