Okay, thank you, everybody, for joining us. My name is Mark Delaney, and I cover Aurora for Goldman Sachs. I'm very pleased to have with us today Chris Urmson, the Co-founder and CEO of the company. Thanks. Thanks for joining us. Thanks for having me. Well, lots for us to dive into the company and the AV industry more broadly, but I thought maybe to start, you can provide a little bit more background on Aurora, what you think is unique about the company relative to some others in the industry. Yeah. So, our company's Aurora. We're on the mission to deliver the benefits of self-driving technology safely, quickly, and broadly. We're building a driver that can operate all kinds of vehicles, but first, we're gonna launch it in trucking. We work with truck manufacturers that make about half the trucks in the US market, so that's PACCAR and Volvo Trucks. And then we work with some of the biggest names in moving freight, so whether it's FedEx, Werner, Schneider, Hirschbach, Uber Freight, and others. So really exciting times. We've shared. We went public a bit over or a bit under two years ago. We've shared publicly a roadmap. Recently, in Q1, we achieved Feature Complete, which means our thing, our truck, does all the things it's supposed to, our driver does all the things it's supposed to. By the end of the year, we hope to have that refined and validated, so we'd be ready to launch if we had a truck platform that was ready to go, and then we expect to commercially launch next year. So things are going really well. We're, you know, getting close to actually having to drive those trucks on the road with nobody in 'em, and, you know, we're continuing to grow the volume of loads we haul today. That's great. Well, I think we were chatting a little bit earlier, and, you know, you made the point you're measuring in months rather than years at this point, which is really exciting. Yeah, even better than quarters. Yeah. Yeah. One of the things that I think is really interesting, the company has a proprietary FMCW LiDAR as part of its hardware kit. Can you talk a little bit more around what that technology does and, and how that perhaps compares with some of the other merchant LiDAR offerings that you may have been able to choose from? Yeah. So, so when we founded the company, one of the things we knew was to drive trucks safely on the road at freeway speeds, we had to see a long way down the road, and as we looked at the offerings that were in market, this was 2017, it seemed like everybody had a LiDAR startup. We looked across them, and we couldn't find one that had the specifications necessary where we had to see 300, 400+ meters down the road to be able to drive safely. Ultimately, we found a company in Bozeman, Montana, that had some amazing technology, and this technology is called frequency modulated continuous wave. And really, the idea here is instead of trying to put a super bright spot of light out in the world and measure how long it takes for it to come back, we send a continuous wave out and self-interfere that, and that allows us to be more immune to noise. It allows us to be more efficient with the energy we put out, so we can see farther with the same amount of energy. And we also instantaneously get the speed, the thing that we're seeing is out in the world, and so that allows us to react more quickly. We've got some videos up on our website, where, you know, we look at what a conventional LiDAR can see and what we're able to see with FirstLight. We're able to see things far enough in advance that we can even complete a lane change around something stopped on the side of the shoulder, for example, before the conventional LiDAR we even see it in the first place. And that matters a lot when we're talking about actually delivering a real product into the world. I think one of the most talked about and written about technologies of the last few months and this year has been AI, and you are no stranger to AI technology and capabilities. You know, as you said on your last conference call, you guys were doing it before it was cool. But maybe talk a little bit more around AI. Very impactful, I think- Yeah - too, as you develop your technology. So, how does AI play into developing your AVs? Yeah. So, so AI is one of the tools in our toolbox. And in building Aurora from day one, it was, let's make sure that we're using the most modern techniques that are available, things like Transformers. But we're also not throwing out, you know, decades, if not longers, work on algorithmic design, whether it's things like Kalman filtering or things like putting in place, hard constraints so that the AI system doesn't extrapolate into a behavior that wouldn't be safe out in the world. And so I think that's been one of the keys that we've had, is not running away, being overenthusiastic with an idea, but making sure we take the essence of it, find the right value to deliver in this kind of safety-critical application, and then utilizing it. So Transformers are one of the technologies that's at the heart of, you know, these large language models, and it's a way to basically have a neuron or a half part of the model look at a sequence of things and be able to predict the next. And in our case, we use that for looking at how traffic is going to interact in the future. And it's an important part of how our combination of motion planning and prediction works together. Really, really interesting stuff. On the last earnings call, you highlighted the release of the Aurora Horizon Beta 7, with advancements in a number of different capabilities, including on-road performance. And you also reiterated some of your targets. And, you know, maybe just talk a little bit more around where the company stands and some of the milestones you hope to reach over the coming months and quarters. Yeah, it's an incredibly exciting time. The company is firing on all cylinders right now, and delivering on what we said. So we have a public roadmap out there that you can find. We hit Feature Complete at the end of Q1. We're working towards Driver Ready at the end of this year, and that means that we validated and have confidence in the performance of our system on our launch lane, so that we'd be willing to let it go without somebody in the truck. And then by the end of next year, we expect to actually have that happen, where we've integrated onto one of our OEM partners' vehicles, and they're actually out there serving customers, moving goods through the world without anybody in the truck. We're also doing things like building the book of business. So in Q2, we shared that we were pulling over 50 loads for customers every week, and we're on a path to try to get to, we anticipate 100 loads per week at the end of the year. So we're growing that and demonstrating the value to our customers and showing the demand that we see out in the market. You know, anything as you think about getting to that driver-ready milestone that's particularly hard to solve from an engineering and technology perspective? There's lots of good, hard work to do, but there's nothing that we look at and say, "Geez, we don't know how to do that." That was really the point of being Feature Complete, is there was at least kind of the first version of everything that we needed in place, integrated, running on the truck. And so now it's about putting in place all the validation. And this is one of the big differences between building kind of normal SaaS software and building safety-critical SaaS software, is with normal software, you kind of run it, it mostly works, maybe you write some tests, and you're like, "Okay, it seems to work. We're good." In our case, it can't just show that it. You know, we don't wanna just look at it and see that it worked a couple of times. We actually want to have conviction that it's going to work across a variety of different conditions. And so that's the work we're doing now, is taking the experiences we have on the road and looking at those things that we see every day. Those are easy to validate because we see them every day, and if something isn't gonna go right, we know about it very quickly. There's things that we see rarely, but we do see, and those we take and amplify and create more instances of them. And so we ask, "What would happen if we were a little bit earlier or a little bit later to the situation? If they were moving a little quicker, a little slower?" And we can test against those and get conviction that we're going to handle them properly. Then there's all the stuff that we're not going to see, right? The stuff that generally results in really significant accidents. There we've taken NHTSA, the National Highway Traffic Safety Administration's taxonomy for how crashes happen, and then we've created variations on those. I think we're at tens of thousands of simulations at this point, where we can look at all of these different types of accidents and make sure that across those, the Aurora Driver is gonna behave the way we'd want to in the situations. One of the measures you have to make sure you're ready is called the Autonomy Readiness Measure, or the ARM, you know, as part of closing your safety case on your- Yeah -on your first lane. Maybe talk a little bit more in detail on that, that specific measure and where you stand. For sure. So for us, it's not about doing a demonstration, right? It, you know, it doesn't matter if we can get somebody out of the truck and run it for 30 minutes one time. What we really want is a product that is safe, reliable, and scalable, and commercially relevant. And so the way we're tracking that is through a thing we call, internally, is through a safety case. You can think of this as the explanation across both the function of the technology and the business, why we believe we can be safe on the road. And what we've done is we've broken that down into all the different chunks of work that need to get done. There's 460-some of them at this point, and for each of them, we've estimated is this a small amount of work, a middle amount of work, and a large amount of work. We've been tracking how much those, what percentage of each of those is complete. Then we roll that up into a measure we call the Autonomy Readiness Measure, which you can think of as how many person-hours of work it was gonna take us to get from Feature Complete to Aurora Driver-ready, and what percentage of those person-hours are done. So at the end of Q2, we shared that we'd done about 65% of the work we needed to do to get between those two. Importantly, this is not one of those kind of measures of, hey, we're at 80% of the number is 20% of the work kind of measure. This is really intended to be a you know, proportional to the actual amount of work that needs to be done. You, you mentioned needing the truck providers to be ready, in order for you to do your own launch, and you commented on PACCAR and Volvo as, you know- Yep your key partners. What's your visibility into them aligning on these time frames that you've articulated? Yeah. So we'll have to ask them to speak to the timelines, 'cause that's their business. But what I can tell you is that we work with them on a daily basis. We work very closely with them. We're helping where we can with defining requirements, helping them solve some of the technical work that they're doing, and of course, they're informing the work we're doing as well. So, like I said, it's a daily engagement with them. You commented on some of the differences of fleets you work with, FedEx, Werner, Uber Freight, Schneider. Maybe talk about some of the things you've learned as you've- Yeah gone through these loads and, and I think it's 50 per week. Yeah, and we look at this as a bidirectional opportunity for bidirectional learning. So for our customers, you know, we're building a technology that we think will have a huge, profound, positive impact on their business, and we want them to begin to get comfortable with it so that we can, you know, plan to ramp. We want them to understand how it works. We understand the limitations, so that as they do need to, you know, change their business logic, that we're positioned, or they're positioned to do that. From our side, it's fascinating to see how these businesses work and understand that, you know, some of them, it really matters that they're driving at night. Some of them are driving during operating hours during the day, right? And it's not necessarily obvious from the outside. Understanding the trailer pool that's out there. You know, you probably know this, but, you know, there's a many to one relationship between tractors and trailers. And the maintenance on those trailers is not always what you might hope it would be. And so how do we make sure that the Aurora Driver is appropriately noticing when trailers are out of spec, but is operating in the conditions that it can, and flagging and saying, "Hey, this trailer shouldn't be on the road- Yeah. -where it shouldn't," and understanding that, that limitation, and, and again, the friction that induces in some situations with the customers. You announced a partnership with, Continental earlier this year, in order to help with your go-to-market, and it's a hardware-as-a-service type of a business model based on miles driven, and I think, targeting 2027. Maybe talk about some of the pros and cons of that relationship- Yeah. and how that can help Aurora to scale. Yeah, this is, you know, I feel incredibly fortunate to have this partnership for Aurora. Continental, as many of you know, is one of the world's leading tier one auto parts manufacturers and technology companies. And for us, we have been looking from early on, how do we build a partnership or how do we build an ecosystem where we don't have to do stuff that we're not good at? Continental knows how to deliver parts at scale. They know how to be part of an OEM supply chain. They understand how to take our technology and help industrialize it, and reduce the cost of delivering it. And so it's a really good relationship for us, and I think from their perspective, the opportunity to work with one of the companies that's leading in this technology and helping move them forward is really valuable. There's also this really exciting part of this deal, which is the hardware as a service part, which means they're investing $350 million upfront to co-develop with us, to set up manufacturing, to put in place financing, to put in place service and support, and then we'll pay them part of our revenue stream to support to provide this hardware to our customers. This is a big endorsement of our approach. You know, if we don't generate a business here, they're not gonna see returns. But when they do, that's an incredible opportunity. And for them, as a tier one, they've been—you know, these tier ones have been looking for decades at how do we get from a should-cost for the number of resistors and transistors in a box to value pricing, you know, service to basically unlock the value of their IP, and this is a first step in that direction for them. So they'll get some experience with that as well, which is exciting. Yeah, I think, you know, the longer-term vision is that Aurora does not own lots of trucks, right? You're the technology provider, and- Yeah ... and Conti plays a role as, as you scale with that. But in the meantime, how do you guys go to market for the first few years? Yeah. So, in the long term, as you said, you know, our expectation is to provide the drivers a service. So, you know, if you're a FedEx, you would order a truck from PACCAR, you'll get it with the Aurora Driver hardware installed in it from the factory, and then you'll have a subscription to Aurora to drive that truck for you, which will cover that additional hardware that's part of the driver. In the near term, we today own and operate trucks. Now, we don't operate them, well, technically, we are a carrier by federal law, but we're not really going out and getting our own business. We're operating for customers in their business, and this is an opportunity for them to, again, learn about the product, get that experience with it. Today, we make the parts. You know, we have small-scale manufacturing that puts the components on it. We do a lot of qualification and testing of them to make sure that they're, you know, designed up to the spec we need. Maybe we can talk a little bit on the competitive landscape. Yeah. There's been quite a few changes with some of your competitors, either shifting focus or unfortunately, ceasing operations recently. Can you talk about who you see the key competitors being for you, and, you know, have you seen any impact from some of the, the change in business- Yeah focus from some of the others in the space? Yeah. It, it's a pretty exciting time to be in automated trucking as Aurora. So we've been keeping our head down and executing over the last several years. And what we see around us is the landscape, the competitive landscape clearing. Two of the public companies that were competitors in the space, Embark and TuSimple. Embark is ultimately sold, you know, for cash on balance sheet, but less than cash on balance sheet. TuSimple appears to be leaving the U.S. market. And so in terms of independent, you know, publicly traded companies, we appear to be the one right now. We know that, Daimler Truck has a subsidiary in Torc, where they're developing the driving technology along with their base platform, and so we'll see where that plays out. In July, they had an Investor Day and talked about their launch timeline and the scale of their business, and it looks like it's about three years behind our timelines. And then finally, Waymo, who we see as an incredibly competent competitor in the space, has backed away from trucking and is focusing on ride-hailing. So as we look at the landscape, there's an awful lot of opportunity for us to go and have an impact. Well, I think, AI and software engineers probably have been in short supply. Have you seen any change in your ability to bring in key talent with some of these changes in the competitive landscape? Absolutely. And again, when we talk about doing what we say and building trust, that doesn't just work with investors and partners. It works with employees as well. And what we've seen over the last couple of years, of course, our stock has been on a bit of a ride, along with pretty much everyone else. And that had one really positive impact in that it was almost a filter for the missionary versus mercenary, right? There was a period of time where self-driving was incredibly sexy. It was a thing to be in, and everyone was going to get rich quick. And some of the folks whose interest was not necessarily in actually making the technology work, but it was like, "Let's go make some money," are now over at Facebook, selling ads or wherever they go. And that's great, and we're really thankful for having them with us, and we appreciate their efforts, but we've now got a core of folks, you know, our team, 1,800 strong, who are about this and about delivering this technology, building value, and seeing this through. And that's incredibly exciting for morale. And then the other part of it is that, you know, we didn't have big layoffs. We've been thoughtful about how we're spending cash. We've invested in areas that matter to us. We've been strategic in that. So I think employees look at that and say, "Okay, this is a place I can rely on." Because, you know, when the Googles and Facebooks of the world are laying off 10,000+ people, and we're just doing what we said, keeping things up and executing well, you know, it's a great place to be. ... a couple financial topics, Okay. I'm hoping to be able to cover. One is on potential pricing for the Aurora Driver, as we're starting to get closer to the commercial launch. Maybe talk about how investors should think through the pricing- Yeah. Does that change at all with the Conti relationship? Yeah, so as we think about pricing for this, we're consistent in this, that we look at value, we think of this as a value-based pricing opportunity. We're providing the equivalent of what a human is doing in the seat, and that's really how you should think about the price that we offer the product at. And in fact, as we look at the opportunity for customers, we see it helping them on the bottom line and the top line. While the direct cost for the driver may be the same, our customers won't have the indirect costs of having to recruit, having to deal with the management overhead of this, all of the bonuses they have to pay to get people to come and work at the company, given the shortage or dearth of drivers in the industry. So we see that as a win. We see an opportunity to improve fuel economy, which is one of the major costs for these businesses, the cost of fuel. We ultimately expect cost of insurance to come down, and we expect the vehicles to be safer, which will help their business as well. And then on the top line, because the Aurora Driver won't be limited to 11 hours a day, like people are, we expect to be able to drive this really valuable asset and generate revenue at maybe twice as much as what they can get with people. And so that seems really exciting, for an industry whose margins are as tight as these are. So as we think about that pricing, the pricing umbrella of human labor is going to be the, you know, basically our competitor or the, you know, support our pricing for, you know, as far out as I can see in the future. Another financial topic has been the balance sheet, and the company recently did a raise of $850 million, bringing total liquidity to $1.6 billion, and the company said it expects this to fund Aurora into 2H 2025. Based on the financial disclosure you provided, you'll need to raise another $800 million-$900 million in order to reach free cash flow positive. Could you discuss what options there could be for the company to further strengthen its balance sheet? Yeah. Well, maybe first I'd start with the fundraise we just did, which was exciting. You know, when we think about raising $850 million for a pre-revenue company in these markets, we're just really appreciative of our partners in this and think it sends a strong signal about what we're doing and how we're doing it well. And when you look at that, the composition of that, it was about $600 million in a private placement and $250 million in an overnight and greenshoe, and that $600 million was about 10 investors, so really concentrated, and I think that, you know, that's just really important. As we think about future fundraise, we'll be opportunistic. In this case, an investor, one of our institutional investors approached us and said they'd like to invest, and, you know, that catalyzed a round. Given the strength of the investor base that came with us, we anticipate, you know, as long as we continue to execute as we're supposed to, that others will be excited about this, or they'll be there again in the future, we hope. Perhaps we can get to some of the financial topics again later in the session if we have time, but I wanted to talk on some other industry questions, and one's on the regulatory landscape. Cool. Maybe speak a little bit around the regulations or perhaps lack thereof in many instances. Do you think there's meaningful regulation, you know, coming that Aurora may need to comply with? Yeah. So this is, I think, one of the parts of our industry that's least understood. The US and Europe have very different regulatory regimes. In Europe, there's a type classification or type certification model, which means there's a box, and your product has to fit in the box, and if you do, and you have a third party certify that, then you can sell it. And so it actually limits innovation in what can be shipped. In the US, there's almost a diametrically opposed point of view on certification. The vehicles have to comply with the Federal Motor Vehicle Safety Standard, and if they do, then you can ship it. So it's really a check-the-box kind of model. So at the federal level, there isn't a constraint on what can or can't be shipped in terms of automated vehicles. Now, there's always an overarching protection that the federal agencies have, which is if you're creating unreasonable risk on the roadway. I apologize for that, then you know, they can pull you off the road. So they have a stick, but you're allowed to bring it out to the market. And so from a federal level, we don't see any real constraint in building and delivering the product. On a state-by-state basis, the vast majority of U.S. states, some 40, some of them, we could launch a truck today if we're confident in the safety of it. Now, we are certainly seeing an uptick in pushback on this, and this last year, we saw 12 bills introduced across the country that were requiring there to be a driver in a driverless truck. Eleven of those were defeated. There's one in California that's continuing to move through the legislative process, and we'll see how that turns out. We're obviously hopeful that California will want to realize the both the safety benefits and the economic benefits of seeing this technology deployed here. That's a helpful context. Inevitably, there are going to be accidents, even with autonomous vehicles, even if it's not the fault of the autonomous vehicle. How is Aurora planning for those sorts of scenarios? Yeah, and this, this is, you know, just a critically important part of actually delivering a real product. And so we, on a quarterly basis, run scenarios where we practice and prepare for this. In fact, I think it was in March, one of our vehicles was in a collision. So as we were driving on the freeway, between Dallas and Houston, somebody fell asleep or nodded off in the vehicle next to us, swerved over, sideswiped one of our trucks, and bounced off. Now, fortunately, this was about as good an accident as you can have between a car and a truck. Everybody walked away. No one was injured. The car had, you know, comparatively minor damage for hitting a truck, and the truck had very little damage. that allowed us to exercise our emergency response procedures in a way that was very low stakes, but real. There we, you know, go through a process of investigating, understanding what happened. We communicate with state and federal officials. We communicate with our partners, both the carrier and our OEM partners, and go through the playbook. At the end of the day, we were able to look at this, and honestly, it showed a demonstration of just how valuable our technology can be in this scenario. So we have on board a driver monitoring system, like many of the carriers have in theirs, and we know when the moment of collision happened. But if you look at the video, all you see is the driver going like this and reacting. So, you know, if there was ever a dispute about the facts, it'd be very hard to provide data to, you know, kind of confirm what happened. In our case, because we have a plethora of cameras and sensors around the vehicle, we are able to look at exactly what happened, and we can see we're driving along in the right lane. This vehicle swerves across two lanes of traffic, bumps into the side of the truck. Our truck, the Aurora Driver immediately understood that it had been in a collision, begins its emergency response procedure. Fraction of a second, the driver is taking control of the vehicle because that's what they're trained to do. We watch the other vehicle bounce off, come to a stop on the side of the road, right? And so we can understand exactly what happened, why it happened. We confirmed that our processes worked, and that the onboard system actually detected the event appropriately. So yeah, we're ready for that. Let's say, hypothetically, there was a problem with the Aurora Driver, or the truck- Yep. It's the truck that causes the accident. Talk through liability and how does Aurora plan- Yeah - for those sorts of situations? I think in this case, liability is generally relatively rational. So, you know, let me not be naive. So first thing that's gonna happen is everyone's gonna get sued, and the people with the most money are probably gonna get sued the most. But the actual structure of liability here is gonna look rational. There's a model for this. It's product liability. And so as we work with partners, we put in place agreements that allow—you know, basically reflect that. We're building a product, it should operate well. If it doesn't do what it said it would, then there's, you know, there's risk to us, exposure risk there, insurance risk to us there. If in contrast, one of our customers, you know, does something that, you know, whacks it with a wrench and breaks something, well, that's probably on them. And if one of our suppliers delivers a part that doesn't meet up to its specification, that was critical and, you know, then it's probably liability on them. But it's just a rational product liability regime is what we expect. Are you partnering with any insurance companies in order to mitigate some of the risk as you do deploy? We continue to have conversations, and of course, we have insurance, but we don't have a detailed partnership there. Okay. What do you think some of the biggest obstacles are that the AV industry is still facing, and how do you expect companies to go about tackling those challenges? I think it's actually a really exciting time for the industry, and we're going to be moving from the challenge of technical viability to business model viability. And this is where I see the real opportunity for what we're doing with Aurora, right? As I look at the trucking opportunity, the unit economics are better, the market is established and there to go and engage, the customers are much less emotional. You know, and what I mean by that is, if you get in an Uber today, and the driver doesn't take the road you expected, or cuts in front of a car that you would have waited for, or maybe worse yet, doesn't take the gap that you would have taken, you get irritated and upset about that. Whereas if you roll a toilet paper in the back of one of our trucks, you don't care, right? And the folks that are paying for that want to make sure the product gets there safely. They want to make sure it gets there on time, and that's it. And so as we think about that, entering that market, where better unit economics, much larger scale, much less emotional customer, much more scalable, we expect technologically, it feels like exactly the right place to go play. You do have aspirations, though, to eventually be in the robotaxi business. Maybe talk about how much the technology can be- Yeah ... utilized across the trucking and, you know, relative to, to robotaxis and how much reuse there may be. Sure. So today, not only are we partnered with PACCAR and Volvo and carriers, we're also partnered with Toyota, the world's largest car manufacturer, and Uber, the world's leading ride-hailing platform. And the technology that we're developing, we designed it to work across different vehicle types. So same software and hardware that's driving our trucks today also drives Toyota Siennas, and we have them on the road in Texas today. We expect to enter the ride-hailing market after the trucking market. We want to be building a business in trucking. We want to be able to generate revenue from that and use that to allow us to enter the ride-hailing market, where we have taken a bunch of the cost out of delivering the product, where we've got a revenue-generating business because those margins are so much tighter in ride hailing. And so we, you know, we see that as the path to enter and then working with Uber, where we're providing some set of trips on it. We don't have to drive all the trips, we don't have to serve everyone. We can serve the trips that are useful to them and to their customers, and then over time, expand the footprint that we support there. But that's after trucking. Any kind of volume or profitability? I mean, do you need to be free cash flow positive or anything like that in order- We're not providing any guidance on that. Yeah. So- Okay. There's just some time after trucking, you think you should... Yeah, we wanna get trucking into the market. We want to be beginning to grow that, and then we'll look at, you know, when's the right time to enter that. I've been looking forward to asking this question. I asked it last year, so we can see if your answers change at all. Uh-oh. The question is: when do you think consumers may be able to buy an AV for personal use that is alpha capable? I think that's a long time away.... Right, and I don't know what I said last year, but, you know, AV technology is really well-suited to fleet businesses. Right, you can increase the utilization of the vehicle. You can, you know, that decreases the cost to the consumer. It allows you to manage and maintain it more effectively, and so my expectation is that will be the way it is for quite some time. Today there's, you know, some products that are perhaps misleadingly labeled, as automated driving, and that's probably not the right way for this technology to come to market. You gave a similar answer last year. Did I? It's gonna be a while, so yeah. Look at that! Okay, not too bad. Well, we're so fortunate to have somebody with Chris's background here with us, and so I do wanna give the audience an opportunity to ask some questions if they have any. Otherwise, I can ask the last couple. Yeah, maybe just wait for the mic, please. My shirt. Everyone... I feel like I'm incorrectly dressed. Blue and- Thanks for taking the time. I was wondering why you think Tesla hasn't adopted LIDAR despite, you know, having maybe a decade head start on data and miles on the road and testing different types of use cases, and at one point, testing LIDAR. Why do you think they didn't go down that path? I'm sure they don't suffer from a lack of talent, and you guys have great talent as well. Yep. So curious why those two approaches diverge. Yeah, I think there's two parts to it. I think the rational part of it is business model, that they have committed very deeply to this idea that they're gonna give the hardware kit away and leverage the data that comes back through machine learning magic to be able to drive the thing. And you can't do that with a hardware, a higher cost parts kit. And so you're kind of stuck in that world at that point. I don't believe it's gonna work out for them, but I kind of understand if you say you're giving it away, and your whole thesis is about maximizing data gathering through customers driving your vehicles around. Got it. I think there's another part of it, which is almost religious, that, you know, people have two eyes, therefore you can do it with cameras, therefore, that's what you should do. That feels like it is honestly part of their argument. It feels like in that case, our cars would all have legs instead of wheels. Right, and we look at it very much from a pragmatic, how do you engineer this thing? And cameras are great, and we have a bunch of cameras on our truck, but they're passive, which means if there isn't light, you can't see. They don't record geometry directly. You have to infer that. They can't see through certain conditions that radars can. So we see that set of complementary sensing modalities enabling a much more robust product, much sooner than, you know, what... You know, ultimately, at some point, I expect some new cameras, new computation, new algorithms will get there, but don't see it happening that soon. Time for one last question. Either it can be from the audience, otherwise I'll go ahead and ask it. One question that I get asked is how to think about the scaling and, you know, we talked about initially, Aurora will own some trucks- Yep … and over time, you're more of a technology provider. But talk about how you think about that scaling, and maybe what does that mean for your CapEx as you think about the next couple of years? Yeah. Yeah, so we intend to be an asset-light business. So this driver-as-a-service model is gonna be how we scale and build the business. Between now and that coming into play in 2026, 2027, we're going to have to own some assets, and we're gonna do that to allow, again, our customers to understand, to learn, to have a lower barrier to entry. But also so that we can maximally learn from those vehicles. It's new technology. It's gonna have things, quirks about it, and we'll be able to figure that out before the customer has to deal with them, and that feels like a benefit to us. In terms of CapEx, it's gonna be a, you know, it's gonna be a relatively modest-sized fleet, so we don't see it being a huge expense over time. Great. Well, unfortunately, we are out of time, so we'll end it there. Chris, thanks very much for joining us. Thank you very much. Thanks for having me.