Aurora Innovation, Inc. (AUR) Earnings Call Transcript & Summary

Thank you, everybody, for joining us. My name is Mark Delaney, and I cover Aurora for Goldman Sachs and very pleased to have with us Chris Urmson, the Co-Founder and CEO of Aurora with us today.

Yes. Well, thrilled to be here. Thanks so much for having us.

Look, you have a lot of stuff to get into in the session. And maybe to start, I'd love to get a little bit more on the background and how you went about establishing Aurora? You co-founded the company after having been at Google previously. So maybe talk to us a little bit more about what led you to go down this journey with Aurora?

Yes. So Aurora is about 6 years old now. We're a company founded with the mission to deliver the benefit of self-driving technology safely, quickly and broadly. And what we're building is the ability for vehicles to drive themselves, a combination of software, hardware and data services that enable vehicles to operate safely. As you said, I was at Google for many years. I helped found what's now Waymo, ran that for a number of years. And in 2016, it was time for me to move on. And I spent some months thinking about what do I want to go do next and spent that time talking with different automotive companies, talking with other sectors, realized that there was this unique opportunity to bring together people with amazing experience in the space and to go build something transformational. And so that's what we've been doing at Aurora. We're focused on how do we do the thing we do best in the world, which is deliver the driving technology and then work with partners to deliver that and scale that in the market. And so for example, today, we work with incredible partners of Volvo and PACCAR, who make trucks. They're about -- make about half of the trucks sold in the U.S. every year. We work with Toyota, the world's leading car manufacturer. We work with Uber, world's leading ride-hailing platform; FedEx, the largest carrier in the U.S. by tractor/trailer count. So we've kind of taken this approach and off to the races with it.

That's great. And maybe you could talk a little bit more on your product, Aurora Driver and how that compares with some of the other offerings that are out there?

Yes. So the Aurora Driver again is the software, which you normally think of as the self-driving stuff. The hardware in the sensors, the computer, the networking on the vehicle and then the off-board data services, I think what we call Beacon, which is the ability to kind of interact and interact with those vehicles and use them in a service and kind of understand how the fleet is behaving. What makes us different? We think there's a few things. So, one, the experience of the people we have, we think is we take them pound for pound against anybody. So we've got an amazing team of folks. We have some really interesting technology under the hood. So one of the really compelling parts of the system is our LiDAR technology, FirstLight. It's FMCW LiDAR. It's a special kind of LiDAR that allows us to see further than others. It allows us to instantly see how fast things are moving. And that's really important when you're working in trucking applications where you have to see a long way down the road to drive safely. Some of the other really novel things that we're doing, we think, are in simulation. So at Google, I helped pioneer the kind of brute force, go drive as many miles as you can on the public roads approach. And in starting Aurora, we realized like, yes, you absolutely do need to be operating on public roads, but you're not going to get enough data from the road to be able to actually have conviction that the thing is going to be safe enough and good enough to drive. And so at Aurora, we've invested in off-line tools for simulation and other off-line data analysis that allow us to test more efficiently, allow us to test more quickly and have higher confidence ultimately that the system is going to be safe enough to launch.

Who do you think are the strongest competitors that you face?

Yes. So right now, we're very much in the early innings. So it's very hard to kind of forecast exactly how the future is going to play out. But we look at Cruise and Waymo, of course. We see those as strong players. The team I have -- used to lead, amazing set of people, good technology. They've been at it the longest. So we have a ton of respect for them. And that's really the top. And Cruise is primarily focused on light vehicles. Waymo also primarily focused on light vehicles. So of course, they do stuff in trucking. At Aurora, we're coming at it a little differently. We're focused on trucking first. While the technology we've built works both for heavy trucks and light vehicles, we see the opportunity in trucking to be just that much bigger than ride-hailing. Trucking today is about a $700 billion market. Ride-hailing is about a $35 billion market, so just massively bigger opportunity. The unit economics are stronger. So for trucking, we think the -- really the long-term competitor for us in trucking is people driving trucks. And they get paid about 3x as much as people driving Ubers. And so the unit economics are much stronger. And if we look at the technology, we think we can scale it much more rapidly by operating on freeways first, right? If you think about a freeway here in California versus freeway in Texas versus a freeway in Minnesota, they're all basically the same. So you crack at one place, we expect to be able to crack it across the nation.

During the last earnings call, the company highlighted some of the advancements that it's made with Aurora Driver. You talked about some progress with things like construction zones and the fault management system. And obviously, you've made a lot of progress so far, but you also laid out a road map, I think you still need to accomplish. Maybe walk us through what are some of those things that are still left to solve before you get into being feature-complete with the Aurora Driver?

So one of the things that's incredibly exciting is we're now at this place where we understand kind of what it's going to take to get across the finish line, right? And the team has been doing amazing work, and we have that visibility. And so we laid out 3 critical milestones between -- over the next couple of years. The first one is what we call feature-complete, and we're working towards that at the end of Q1. And feature-complete means that at that point, the Aurora Driver does all of the things we think it will need to do to ultimately be a product on our initial launch areas. And so that will be a major milestone for us. And we've outlined some of the technical things that we're going to do between now and then, and we've got a couple of them where we've shared that we'll be done with by the end of this quarter. From there, we expect the Aurora Driver ready to happen at the end of next year. And that means that if we had a truck where the autonomy -- or sorry, where the base truck had the redundancies and interface is necessary, that we would be ready to go, that the Aurora Driver would be safe enough to put on the road and operate and start hauling loads without humans onboard. And then at commercialization in the back half of '24, and that's really dependent on having those autonomy-enabled truck platforms in place. And so those are kind of the 3 big milestones you'd be looking for from us over the next couple of years.

When you think about the things that are within your control and having the product ready by the end of 2023, if the trucks were there, how are you handicapping for hard-to-solve technical challenges, right? I mean, we've seen as industry observers and investors, a lot of delays in autonomous technology historically. I imagine you're trying to handicap for some of these things in your forecast, but maybe talk about the visibility you have.

Absolutely and right. And this is why we're now at a place where we feel confident to carefully share these milestones is because we think that we understand what those are. We understand how we're going to tackle them and that we feel in a good place to be able to accomplish them. And right, we've been building the technology for almost 6 years now. We spent the first 4.5, 5 years of the company really focused on that core underlying technology that enables it. And now we're in the productization, get it out the door and start actually doing useful things the way we've been hoping to.

No, it's pretty amazing. And even the product as of 6 months ago, right, when you had your Investor Day down in Texas, and I was in one of your trucks and the demo-rides and 0 interventions. So -- and that's already working really well. But I know you've got to get every single one of those edge cases solved before you are ready to deliver.

Yes. It really is very exciting, right? So we are on the road today, hauling loads for our partners with FedEx, with Uber Freight, with Schneider, with Werner, right? We're out there doing useful things, but we have people on board. And for us, the ultimate threshold for being able to take the people off is to complete our safety case. And we're one of the first companies, I think, maybe the only company that's actually shared their safety case framework on how we're going to actually get from. I think that works really well for an experience that we can give analysts to one that people can trust every day on the road with hauling their goods and around other people driving there.

Segues well into one of the other things I wanted to ask you about those customer trials that are underway, what are you learning and what's some of the feedback you're getting?

Yes. We're -- like for us, the intent of our pilot programs has been to look at the different segments of the truck and make sure that we are working with partners that represent each of them because less than truckload is different than a captive fleet, right? And so we want to understand the different experiences that these customers have. For us, it's been fascinating, right? So we've understood how the complexities of trailer management and seeing the fact that while DOT has specifications for the trailers, they don't always meet that. And so there's an opportunity for us to kind of help our partners improve the quality and safety of their fleet by having a closer look because as we're putting the autonomy -- autonomously driving truck in front of it, we're trying to make sure all of these things are in play. The challenges with some of our partners who operate primarily off hours versus partners who operate during what we would normally think of as business hours and the opportunities that creates for time multiplexing terminals and whatnot, like there's just kind of a lot of interesting operational learnings we're getting from it.

You mentioned on your last earnings call that the autonomous trucks won't be ready until 2024, and so the commercial ramp-up is sort of that late '24 time frame. Why is it taking that long for the trucks to be ready?

Yes. So first, we have an amazing relationships with our partners, Volvo and PACCAR on this, and we work closely with them. We're engaged with them on a daily basis and have regular senior executive engagement with them. Building a vehicle program, one of the highest risk parts of that program is supplier selection. So you're developing some new components for the truck or for your car. And there's a negotiation that happens between the OEM and the Tier 1. And there, they're trying to align on requirements, they're trying to align on timing, they're trying to align on economics. And so our partners were going through that process for some of the major components in the trucks, things like the redundant steering system, the redundant braking system. We now understand that those components have been awarded. And so that's a major de-risking event in the program because at that point, it's in the execution mode. And so it's not that there isn't risk in those programs still, but we now -- we have increased confidence in the timing of those programs because this major hurdle has been overcome.

Yes. Do you think if supply chain constraints were to ease, is there any chance that it could happen sooner than what they've communicated or is this sort of the best-case scenario, but one you now have a good line of sight into?

Yes. So we'll continue to work to do whatever we can to pull things forward. But I think at this point, it's like they have these partnerships, they have these alignments and schedules and they're going to be working towards those.

When you talk about how adaptable your Aurora Driver is between trucks and light vehicles because you're trying to do both, but what are some of the similarities and differences? You're doing trucking first, but what would you have to do to make it applicable for robotaxis?

Yes. So this has been part of the way we've architected the Aurora Driver from day 1 is that it really is a common platform that works for big trucks and for light vehicles. And the intuition you can have for this is that, well, first, if you want to become a truck driver, you have to get a commercial or a light vehicle license first, right, what we call a driver's license. You can't just get a truck driver's license. So there is -- obviously, there's some commonality there. From a technical point of view, the hard problems in self-driving, it's not like how do I talk to the steering wheel and turn it. It's how do I understand the world around the vehicle, how do I anticipate the behavior of the vehicles in response to what I'm going to be doing over the next few moments and then take action that factors that into it. And so that core is common to both of them. And in fact, for our system, the sensors and the vast majority of the software is common for both applications. Trucks are bigger than cars. That's the difference, right? Class 8 trucks, they bend in the middle. That's the difference. But the core of it is so much the same. And for us, because we understood from day 1, we wanted a driver that was going to operate in both of these, we haven't kind of cut corners in places to unduly simplify the product. And so it's very applicable. And we showcased that in Q1 of this year when we took the system that had been driving our Peterbilt trucks and have it on our Toyota Sienna fleets and up and operating.

You mentioned earlier in our conversation the FMCW LiDAR that you have, FirstLight and it sees really far out, right? It's super important for a truck because it needs a longer amount of time to come to a complete stop. Is that still an advantage in light vehicles? Do you need that kind of visibility?

Yes, it really is. So you don't need the same distance that you do with a big truck, right? The truck has more inertia. You're going to slow down a little more generally because of that. But even for light vehicles, there's big advantages. So first, when you're operating on the freeway, the farther you can look ahead, we've all been through the driver [indiscernible] I hope. And you talk kind of look down the freeway, you don't kind of look at the bumper of the vehicle in front of you. And so the FirstLight LiDAR gives us the advantage for that. The other big advantage that we talk about, but it's really exciting in urban environments is this software measurement. So this is the fact that for each point measurement that we get back from the LiDAR, we get the speed of the thing there. And this is a really important signal when you're trying to segment, say, a person walking next to a hedge or a person walking next to a road, trees. And in fact, when the moment when I was like we need to bring this technology in-house, I was out with the team at Blackmore that we brought in, and we were driving down a road, but I've looked at LiDAR data for decades at this point and pretty good at it. And you're looking at the screen, and there's like these 3 little dots right next to all these other dots and they were colored red. And if I'd just been looking at those dots, there's nothing I would have known. But I look at the dots and like, okay, well, that means it's moving. That means that's a person, right? And it stood out like day from night and having understood the complexities of machine-learning, that level of signal to be able to segment these 3 dots that just look like any other bush, but are actually a person going for a jog was a magical moment. And that's the kind of thing that will be an advantage even in light vehicles in urban environments.

We spoke already about the time line for truck commercialization in late 2024. What should we be thinking about at this point for robotaxis?

Yes. So robotaxis will follow trucking. We're not providing guidance on exactly when they'll follow. But we see them starting in truck-like environments, so things where you're driving primarily on the freeway, think of it driving from an airport down the freeway to a hotel or to a convention center. And then that will be part of the Uber network, and that relationship with Uber and the partnership there is going to be, we think, a real advantage in bringing this to market.

Are you concerned at all about being late in robotaxis, right? You've got Cruise in small-scale deployments, but they've deployed. We see Waymo around San Francisco where the conference is taking place. If it's a couple of years before you're out there, is that going to be something that's going to be tough to catch up?

Yes. We think we're very much in the early stages of this. And we think getting it right and getting it to market quickly is going to matter. But the scale of this industry is gigantic, right? Transportation is a $2 trillion industry in the U.S. or $3 trillion, something like that. And so when we enter a market, we're going to be entering it a little differently than the Cruise and Waymo folks. They're focusing on low-speed driving in city centers. We're going to be focusing on freeway driving. And so likely, we will enter different markets because the markets that are really good for urban dense driving are probably not the ones that are more kind of sprawly suburban freeway-centric trips. So we see kind of asymmetric entry. And then just a long time, right? This is not a thing where everyone is going to have a self-driving car in a week, right? Just the -- if you think of the turnover rate of vehicles, the average fleet vehicle or the average vehicle in the U.S. fleet is about 15 years old. So that meant if every self-driving car that was -- or every car that was made today was self-driving, it would take 15 years to get to half of them, right? So we get there. We're going to get there well. We're going to get there with the strength of partners that allows us to scale. And we think that all sets us up to be hugely successful. And I'll be honest, though, like if we only win the trucking market, we'll be doing okay. And our investors will do very well through that.

I think you've been, at least from my perspective, very thoughtful around how you want to scale once the technology is ready, both in the robotaxi space with some partnerships and then also about how you go state by state and with some of the geographic considerations for trucking. Maybe elaborate a little bit more on how you're going to partner in robotaxis with the Ubers of the world and then only doing the route that makes sense for you because I think it's pretty interesting.

Yes. So as we've built our models of how we expand and grow the business, we've not just kind of looked and said, okay, well, year 1, we're going to be 0.001% of miles traveled and the year 2 will be 0.01% and then 0.1%. We've actually looked at the kind of the demand in different markets and kind of aligned to that over time. And so for us, it's entering markets under the Uber umbrella or under the Lyft umbrella, right? We can partner with both over time. Today, we're partnered with Uber, and that's a strong partnership in the next roughly decade. And then we'll go in, we'll enter first, picking up these freeway trips. And then as we advance the technology by understanding where the demand is, we'll be able to pick off the next set of requirements. So maybe first, we're on the freeways and then we're like, okay, well, if we do boulevards, that will unlock another 5% of trips. So we can add the technologies and capabilities to drive boulevards. And then we can pick out from there using this unique data access -- asset we have from Uber, okay, what's the next feature we should add and do that in a super capital-efficient way. We think that's going to be a very powerful advantage.

And then likewise, on the trucking side, I think you've been pretty thoughtful around how you're going to build out just starting some select routes. But how does that scale?

Yes, that's -- we take a very similar approach to that. So we're going to start in Texas. And then we're going to build out first likely across the Southern Sunbelt routes, right? These are routes where the weather is better, but where there's a very high density of traffic, right? And there's a reason why there's a Southern freight belt is because driving a truck in bad weather is difficult. And so truckers prefer these routes, we'll be able to operate on them. And then over time, we'll expand out. And again, similarly, we didn't just kind of look at freight miles and call it a percentage of it. We said, okay, if we're going to pick these nominal routes to begin with, and we can -- we expect to be able to take and support our customers with this fraction of the trips on them, what does that translate to? How do we expect that to grow as we unlock representative routes out from there, what's the demand that that unlocks and grown that kind of thoughtfully over time?

There's surprisingly little regulation in the United States around autonomous driving. I think when these rules were written decades ago, nobody would have imagined that we're going to be seeing the kinds of things that you're working on. But talk to me about how you're engaging with regulators and how do you think some of these regulations could evolve?

Yes. So from day 1, we've known that it was important to be partners to the regulatory bodies, right, that a lot of the Silicon Valley model is like keep them as far away as possible. But in practice, that just doesn't work in a regulated industry. And so we work both at the state and federal level and locally, in fact, with folks because regulation as many of you are aware, is tiered. So at the federal level, they are responsible for the safety of the vehicle. At the state level, they're responsible for the safe operation of the vehicle, and then local governments control access to roadways and whatnot. And so we've engaged with the Department of Transportation, both at NHTSA, National Highway Transportation Safety Administration, but also FMCSA, the Federal Motor Carrier Safety Administration, and help them understand what we're building, why we're building, what the opportunities are, answer their questions. We've had the administrator -- well, former administrator now of NHTSA in the truck in Texas, got the experience, understood and helped to educate his staff. We've had similar engagements with FMCSA as well.

And when you talk to potential customers, do they feel like they need to see more regulatory certainty before they'd be willing to use the product?

So we have regulatory certainty today. right, that in the U.S., if we had a safe truck, we could put it on the road, if we were confident in the safety of it. And then the various departments of transportation have sticks they can use, right? They think it's unreasonably unsafe or doesn't meet the reasonable safety goals, more positive way to put that, then they can intervene, right? And they can recall you, they can have state police pull your vehicles over. And so there's a lot of opportunity for them to stop it. But today, because of the relationships we build, because of the safety approach we're taking, right, we have confidence that we will be able to operate these vehicles. Now there are some states, California is notably one of them, where we can operate light vehicles autonomously but not heavy trucks. We're working with state DOT to kind of help them think through regulations to enable this.

Yes. Unfortunately, accidents are just part of this world. And at some point, it will happen to Aurora. How do you respond when an accident eventually does happen?

Yes, absolutely. So collisions are going to happen. We're not the only driver on the road. There's going to be other people and people make mistakes, and we're fallible, right? We're going to do our best to make a system that's safe and out there, but it will have some flaws in it, and we'll work to fix those over time. So what will happen is, fortunately, while people have a lot of thoughts about this, it's actually pretty well tread and we have a really robust legal system in the U.S. So there's going to be an accident or a collision, there's going to be an investigation that happens at that time. We're going to have an unprecedented understanding of the event that happened because unlike to human drivers crashing together and you have kind of their flawed memories of what happened and differences of opinions of it, we will have hard back data of, okay, we're driving down the road, and this car just side-swiped into us and out of nowhere, right? And so that data will help us understand first, is there something we should be doing better and improve it, but it will also help us and our insurers figure out how hard to fight, right, and to address this and take some of the ambiguity out of the process will make it less costly, we expect. Over time, what we see is an incredible opportunity to improve safety and to provide insurance carriers with much better visibility into the risks and thus, that will allow the insurance market to be dramatically more efficient, we expect.

And so closer related to that is societal buy-in and I had the chance to do one of the Cruise rides on Monday evening with a number of investors and I was telling family members and friends and a couple of people said, "Oh, did you feel safe?" It didn't even really occur to me because I know how much you guys are working on this and all the precautions you're taking. But you get those kind of questions and so it kind of ties in with the accident side of it. But how do you build that societal trust?

Yes. So I think the first is that we're actually in a really good place with this. You'll see every year, AAA or somebody else who come out with a study that says 60% of Americans are anxious about riding in self-driving vehicles. What that means, though, is that 40% of Americans aren't. And that's like 120 million people. And that's not a bad starting point for a product, right? I think if you've done a survey about how many people want a brick of glass in their pocket that allows them to look at the Internet, right, before the iPhone came out, the answer would have been basically 0. And so I think that we're actually in a really good place. At the same time, I think it's on us to help educate and inform. It's one of the reasons why Aurora's been as transparent as we have about how we do safety. We've kind of helped found PAVE, which is a nonprofit that's out there to provide education and inform policymakers and the public about the technology. That's why we share as much as we do, and we can. The other thing, though, is that much like you, once people experience it, it shifts, right? And so I think as we see others in the space give people that experience, right, I think it will become easier. And my intuition for this is that we are all used to riding in vehicles that somebody else drives, right? I took an Uber here today, sat in the back and looked to my phone for the whole time, right, kind of forgetting the amazing thing that was this car driving around me. And I think the same happens to people riding in self-driving vehicles. A number of times, I've given people demo rides. And for the first 5 minutes, they're like, my goodness, this is crazy. And then 5 minutes in, they're like, is that all it does? And then 10 minutes in, they're on their phone, right? And so I think that, that -- like there's a normality to it that very quickly creeps in.

Maybe a couple of financial topics, if I could. Balance sheet, I think, $1.4 billion of cash at the end of last quarter. I believe the company said it expects this to last until mid-2024 at your current projected cash use rate. But that's before you plan to commercially scale up. And so potentially, we would need to take steps to strengthen the balance sheet. Talk us through some of those options and any potential avenues for capital raising.

Yes. So we were clear when we went public that we were going to need to raise capital. We kind of had confidence that we were going to be making material progress that would show and kind of decrease the risk to investors and they would be excited to be part of it. Obviously, we didn't anticipate what's happened over the last 8 months. I don't think anyone did. So as a company, one, we're incredibly excited to be in the financial position we are, right, that we have this runway that we expect markets will stabilize and kind of get back to more normal, and we'll continue to make amazing progress. And things will work themselves out and the avenues available to us. By sharing the road map that we did, we wanted to provide investors with a very clear set of kind of milestones and goalpost they could look and see how we do it. And if we're executing well and we look at the opportunity of the $700 billion space, we look at this kind of financial trench that's happened that we expect will make it very difficult for subscale or other players in the industry to actually raise capital. And we think this actually strengthens our position over time. And so as we look to the future where we don't have any fixed plans to raise capital, where internally we're looking at what we can do to self-solve and extend runway by focusing on getting the first product and trucking to market in a more capital-rich environment, we would do more in parallel. And we're going to do a little more in sequence with light vehicles versus trucking. We're looking at things that were nice to have and say, we really need that, right? And let's find these places where we can be more capital efficient. It's been really actually useful as an organization to have this pressure to go through and look at where do we have organization duplication, where do we have lack of alignments and where can we simplify that to create more efficiency in our organization. So we'll take the silver lining on this one.

Yes. You've got partnerships and pilots underway with some well-funded companies to get a lot of economic value out of your product. How possible is it to get some prefunding of truck purchases or prepaid for trips done from some of these well-capitalized partners?

Yes, it is not yet a conversation we've had. Today, we're in a place with partners where we have non-binding indications of interest that exceed the demand, and I want to stress those are non-binding, but it gives us confidence in this kind of if we build it, and they'll be there. And so as we start to get closer to deployment, we'll start to have some of those conversations. But right now, we're -- if we look at our OEM partners, for example, unlike most, if not all of our competitors, we're not actually paying them to do the work, right? They're investors in the company. They're doing the work because they see the opportunity and alignment with us. And we think that strength of partnership, right, that shared vision for what we're doing together is really powerful. And it's a much better motivator than kind of a mercenary relationship.

I've been looking forward to asking you a couple of questions around industry progression more broadly given your experience over the number of years, you've been involved with the AV development.

It's depressingly long list of things.

This is one of the harder questions that people have to think through, but when do you think AVs will be available for personal ownership?

Yes. So I don't know. And part of that is because I don't think that's the right commercialization model, right, that I think that the biggest opportunity for AVs is to create a shared infrastructure for people. I think it is a better application or a better opportunity for society, and I think it's a better economic opportunity. And so I think that we're already seeing fleet operations happening, whether it's us, with our trucks and through our partners' networks or whether it is the Cruise and Waymos of the world out there serving folks. And I expect that is going to be the right model for the foreseeable future. Does it enter high-end vehicles? Maybe. But these products are going to come to market in a geo-located way, the real autonomy systems, the truly Level 4 autonomy systems. And so I think that makes it more complicated for it to be a personally owned product.

That's helpful. And so I guess then by the -- to extend that conversation further, right, if I can, another industry topic is, okay, AVs, what does that mean for the personal market, right? Does it significantly reduce the amount of vehicles sold every year because perhaps you can get the cost per mile to be significantly less than the cost of owning the vehicle, which is today, using an Uber or Lyft is actually quite a bit more expensive per mile than what most Americans pay if they own their vehicle. But do you see an intersection there? And kind of what does that mean for the market long-term?

Yes. I think that's exactly what will happen is that over time, the AVs will enable shared vehicles to be lower cost than car ownership, right? And you talked about cost per mile in terms of operating costs, but you see this shift already happening in places like San Francisco, more so in New York where people don't own cars in the same way that they do in lower density cities, right, because the total cost of them, the cost of parking, the cost of the inconvenience of driving that car, being stuck in traffic, right, and then the cost of car on top it, it doesn't -- the value to them is less than that cost. And so they take town cars, they take Uber, Lyft, they take the subway. And so if we bring the cost down and improve the quality of public options or shared options, then I think that is what will happen in certainly in urban areas over the coming decades.

I do want to give the number of people here in the audience opportunities to ask some questions. So if anyone wants to take the opportunity to ask Chris a question or two? Yes, we've got one right up here in the front. We got a mic coming, if you don't mind waiting for a second.

How much compute power do you need in the vehicle? And how much do you need for your AI training type of vehicle? And where are you sourcing that from?

Yes. I don't think we're sharing those details, but a lot on the vehicle. I think more than -- if you hear a lot of the Tier 1s, they're thinking about these relatively large SoCs that they use for driver assistance systems. What we see for autonomy is dramatically more than that, right? We're using high-end GPUs and processors. And we think that's kind of critical to make this work. Off-board, we use -- is it fair to -- yes, so I think cloud computing is our second largest cost behind people. I think that's -- so we spend a lot in the cloud to do the training and development work here.

Maybe we could continue on our discussion around costs are coming down over time. And how do you think about pricing in the Aurora Driver?

Yes. So for pricing for the Aurora Driver, again, we think the biggest competitor is going to be human labor and that that's going to create a pricing umbrella for the next decade plus. And so as we think about pricing the Aurora Driver, we see the product we're delivering is a premium over human driving, right? The asset, the truck will be able to operate more hours a day because it won't have hours of service limitations. It has this kind of superpower of being able to see all the way around at all times. It doesn't get distracted. It doesn't have -- it doesn't get annoyed about the stupid Toyota that cut them off. I shouldn't say Toyota the stupid. I don't know. Tesla, they cut them off. And so it's going to -- and it's not rushing to get home to see family, right, and because the driver has been stuck away from them for days at a time. And so that means that we can see better safety, better fuel economy, more utilization of the assets. So we see that as a huge opportunity for our carrier partners.

Do you want to talk as well on the manufacturing approach, right? Because I mean, obviously, you're not making your own trucks, but there's a lot of compute hardware going on to these vehicles centers, some of which are your own proprietary centers. So what are you doing yourselves? What are you using for partners?

Yes. So today, because of the scales we do and because we want the rapid iteration of being able to close the loop between the engineering team and production, we develop and build most of the hardware in-house. We use contract manufacturers, of course, for the boards and whatnot. Over time, we will shift this out of house, right? We'll continue to own the core IP where we think it's strategic. And then -- or we expect to continue to own the core IP where we think it will be strategic. And then we'll work with contract manufacturers, Tier 1s to actually take it to scale and inject that into the supply chain for our OEM partners because we -- what's actually really exciting about the autonomy of truck work that we're doing with Volvo and PACCARs, these are trucks that are not going to be prototypes they're building. These are things that can come off the line of scale production. And so we ultimately want the Aurora Driver to be one of the things that gets plugged onto that truck on the line, and we're working with them in those conversations today.

I wanted to close out with a more open-ended big picture question for you. And feel free to answer with anything that you can think of about the industry, your company, even just something of personal interest to you. And so what is something that you think could surprise investors or industry observers over the next year?

Yes. Do I think that the pendulum has swung over the last 5 years? So 5 years ago, self-driving vehicles were done, they're going to be on the road tomorrow, carmakers were going to be dead, right? And that was 5 years ago. And today, I think the sentiments over -- maybe up until the last couple of weeks, the sentiment was like, "Geez, I don't know, if the self-driving thing is ever going to happen. Are they going to get there? Boy, it's hard. A couple of players are having trouble, it's going slow." And I think both of those sentiments have been wrong, right, that we were clearly not going to be there tomorrow. And this is happening for sure, right? And so I think that over the next year, we're going to see more signal in that direction. And the pendulum is going to swing back in the -- holy c***, we have technology and a couple of companies that are going to transform one of the largest sectors of the economy, and that's an incredible opportunity, right? And if you think about major transformational moments, whether it's in communication, like the printing press, the telegram, the radio or worldwide satellite coverage to the Internet, to the smartphone, you think about transportation from the horse to the steam engine, to the internal combustion engine. This is kind of the next big step there. And we're on the cusp. And so I think over the next year or next 2 years, we're going to see like this kind of come out of the valley of disillusionment on this, and we're like holy c***, this is real and it's happening and it's going to be a big deal.

Okay. I think that's a great place to end it. So Chris, I really appreciate you speaking with us today.

My pleasure. Thanks so much for having me.