Rivian Automotive, Inc. (RIVN) Earnings Call Transcript & Summary

All right. Welcome, everyone. We're about to kick off the second Annual UBS Auto & Auto Tech Conference. Really pleased all of you can join us here today in person and online. And to kick the conference on, super excited to have with us Rivian. We have Founder and CEO, RJ Scaringe with us up on stage. Got a couple of their beautiful vehicles outside. So please take a look at those as well. So RJ, thanks for joining in. I guess just to kick things off, the world seems as dynamic as ever. Whether it's policy or macro and those two are obviously interrelated. Can you give us a sense for how you manage that volatility and uncertainty inside the company? And sort of really, what have you learned over the past year? And what has sort of changed to sort of help better prepare you for the future in this ever-changing world?

Yes. I mean we've -- it's funny you asked this. We were joking internally when we launched R1 in 2021 -- end of 2021, it was surrounded by, of course, the pandemic and the lead up to that, we were trying to build a plant for -- it was our first time launching a product of this scale. And so the process of bringing up a supply chain, launching a manufacturing plant and doing that in the context of COVID was just entirely unexpected and very difficult. We then launched an update to the R1 product, which we call Gen 2. And as we're getting ready to do that, we had this massive supply chain crisis. And so now we're getting ready to launch R2, our next product, and we have a major trade challenge. So we were joking and saying every time we launch a product, it's sort of a precursor to some sort of big global disruption.

Like a leading indicator.

Yes. But I mean I say that, but jokes aside, we've actually learned a lot as a company having grown and scaled through some of these really big challenging times. Learning to run a manufacturing facility where we couldn't have people working on site, to then trying to scale production with the supply chain crisis and semiconductor crisis that we experienced in 2022 and 2023. And so we've actually have a lot of robustness in our supply chain processes and our experience in running SWAT team style daily stand-ups. And so what we've been doing now for the last many months, I guess, I should say, is we have a daily meeting of senior leaders, where we talk about what's changed in the last 24 hours and what that tactically means for us. And never did I ever think that we'd be on a daily basis talking about supplier health, supplier situations and trade, but it is extremely dynamic. And the challenges we're making as a company decisions. We typically make these decisions with time scales measured in 5 to 10 years. So these big CapEx decisions or production location decisions. And the frequency for which the environment changes and therefore, the calculus changes on how you might make a decision is just -- it's in territory. And so one of our advantages, we are a really agile company, and we've built a lot of muscle around doing that, but it's a very unique way of operating.

I guess just on that sort of day-to-day noise, obviously, there's a couple of policy element that I think the administration is looking at, that can squarely impact your business, whether that's EV and emissions related stuff here, obviously, sort of trade. I guess just like bigger picture before we dive into some of those topics, like what is your view on the goals of the administration? And how do you think sort of that can impact Rivian? And secondarily, like what efforts are you or the company doing in D.C. or through other areas to sort of help educate the administration about maybe some of these challenges or the negative repercussion of some of them?

Yes. I mean it's a great question. I think it is -- we have to be responsive to what's happening real time and be thoughtful around that. But the important question and something we spend a lot of time thinking about, what are the things we can plan around? And if we try to take all the signals ups, down, left, right and interpret that noise into a more steady-state signal, what are the things we can pull out of it? And so the first is, in this, I'd say it's true whether we're in a Republican-led environment or a democratic-led environment, which is there's a focus on technology being in the United States and manufacturing being in the United States. And that's clearly a shift, and it's a shift in the United States, but it's also just -- it's a global change. We see a lot of large economies shifting to want to create more jobs, create more technology or create more manufacturing in their home environment. And so that fortunately actually aligns really well with our overall strategy. We're a U.S. company. We have well in excess of 15,000 employees here in the United States, produce our vehicles in that states, and so we're incredibly vertically integrated on our technology and what we build here. So we're actually very aligned with the administration in terms of administration's objectives. Our next plant beyond what we have in our initial facility in Illinois -- in Georgia. It's helpful that that's an important swing state. So we're politically in perhaps one of the best states one could hope for. It's a great state for doing business. I have a great relationship with the Governor. And so it's a great place to be scaling the next level of volume for us as a business. But I think above and beyond all that is, looking at what's happening from a customer point of view. Customers want products that are really exciting. They're going to want products that hit like a value proposition, that's unique and compelling. And our job is to work through all the noise, all the complexities to deliver on that. And so I couldn't be more excited about what's coming with our next generation of products with this R2 and we have a follow-on product, which we call R3, both the small SUV and a small crossover. And they are exceptionally -- I think they're -- the opportunity they have in terms of expanding volume for us as a business is really exciting.

And then were talking a little bit earlier about some of the maybe near-term challenges that have come up, whether it's rare earth or additional steel tax, like I think you were mentioning like some of the stuff is just difficult to sort of get around in the near-term. So I guess, what kind of role are you trying to sort of play in educating policy here?

Yes, yes. I mean we, along with a number of other car companies and folks we have great relationships with. We have a huge partnership with Volkswagen Group. They're an investor, and we have a large -- we have a $5.8 billion software licensing deal that we did as part of a joint venture with them. So it's been helpful to have them as a thought partner here. But the time we're spending in D.C. is really focusing on helping the administration understand the nature of more -- like very complex supply chains. And I think often, when we think about these things and they get written up in media in their headlines, it's as if we're talking about an economy that is buying and selling or trading coffee cups and T-shirts, like these really simple products and simple supply chains where it's low CapEx and relatively straightforward to move production around. But on something like a car. We have hundreds of suppliers who in turn have thousands of suppliers, who in turn have tens of thousands of suppliers, who in turn have hundreds of thousands of suppliers. And the ripple of these changes in policy is very, very hard for these systems to adapt-to immediately. And a good way to think about this is just practically go through that change. So imagine how we react when tariffs increase. So of course, my job and the team of our procurement -- or the job of our procurement team is to go to all of our suppliers and fight like hell to say, well, yes, we understand tariffs have gone up, but we negotiated a price that protect us against those tariffs and to hold to that and say, look, yes, your costs have gone up, but we have a fixed price that regardless of what happens with trade, it was -- that's how we negotiate it. And so we -- maybe in a few cases, we'll bend on it. But generally, we push really hard to maintain the pricing that we negotiate with suppliers. Of course, they don't like it, but they then turn around and do the exact same thing to their Tier 2s. They go to their Tier 2s and say, look, we were sourcing this for you. This is what we're going to do. We can make a 5% adjustment, but you're going to have to eat a lot of this, you're going to have to find ways to drive efficiency in your business and they push it to the Tier 2s. Tier 2s don't like it, but they say, okay, fine, we're going to have to deal with this. Tier 2 turn around, they go to their Tier 3s, and they do the exact same thing. And in the end, it gets to a very large number of very small companies, like companies like these are often sub-$50 million a year companies. They're the Tier 3, Tier 4 suppliers sort of the end of the supply chain. And they don't have the financial capacity to absorb it, but the last in the line. And I think this is not appreciated of how the dynamics of these changes get rolled out and what they do to a very large number of businesses across the United States. And so I think the messaging we're providing along with a lot of other manufacturers, along with a lot of manufacturing associations that represents these small businesses across the U.S. are starting to recognize that this is a dynamic that's going to hurt, if we have these rapid changes and don't have ways to ramp into them, it's going to hurt them. And it can be little things like I talked about like, we buy a tail light from a tillite supplier. It's a U.S. sourced part. It's 100% U.S. sourced. That tillite has components they buy from Tier 2s, Tier 2s buy from Tier 3s. Like you move down to a Tier 5, they're providing plastic pellets that go into making a plastic part that goes inside a plastic assembly. And that pellet supplier buys an additive that gets processed in China, and that represents 20% of their cost. That company can't exist if their cost structure goes up by 150%. And that's just -- again, it's my point of like the economy is not running and it's not -- we're not dealing with making T-shirts and coffee cups. It's much more complex, much more nuanced and much more layered. And so recognizing that, there is a desire to shift a lot of things to the United States. There are certain things that we don't have as a country, either the capability or the desire to house. And so a good one here to think about is, which is maybe the most complex because it's not even cost is rare earth metals. When I say it's not even cost, it's a question of can we get those because China is taking -- as the relationships have become increasingly antagonistic, they got to a point where China put in place export controls around heavy rare earth metals. And so a quick Google search would tell you, well, heavy rare earths are not that rare, and you can find them in lots of places. But then the follow-up Google search should be, well where do they get processed? And almost all of the processing of these rare earth metals is in China. And so take like dysprosium, which is used in a lot of power applications like motors, we're not going to build a dysprosium processing infrastructure in the United States. And if we do it, it will take a lot of time and it would take significant changes to our regulatory environment in terms of allowing those types of facilities to exist here. So these are like these big ripples that we need to contemplate. So we've spent a lot of time on these topics. And I'd say the administration has been really receptive. You've seen how they've responded and moved -- changed tone on certain topics. But this is -- it is a 24-hour a day full court press, put a giant team on this. This...

Bigger than you thought a couple of years ago.

This is in the category of hard work.

Let's move on a little bit to -- I do want to get to the R2 and some of the exciting stuff you're working in the future. But just in the very near term, one of the things you talked about on one of the past earnings calls is maybe a little bit more hesitance on the consumer's front to survive bigger ticket items. You do have the overhang here of potentially sort of in consumer EV credit going away or at least from a leasing perspective. So I just want to understand how you're sort of thinking about -- or what you're seeing right now on demand in real time? And then what you think that could mean for demand over the balance of the year? Is there even a potential maybe you get some pull-forward, people taking advantage of the credit that might expire? And then what does that mean for '26?

It's been as chaotic as it seemed in the macro, we see it show up also on the demand side. We'll have a lot of volatility around how consumers are behaving based upon -- you said it, but what's going to happen with consumer-facing credits, perspective on what's going to happen with interest rate, perspective of what's going to happen with overall health of the economy. And today, we have our flagship products. So it's our R1, which is the R1S and the R1T. It's a sibling set of products, an SUV and a truck. And they're high price, it's an ASP. You can see in our financials, it's about $90,000 ASP, but it's very popular. So the R1S is the best-selling electric SUV over $70,000 by a pretty significant degree. Of course, it depends on when you're looking, at which month, but call it, 35% market share, which is remarkable. But the 35% market share is of a relatively narrow TAM because of the price point. And so our hope is that we can continue to maintain that market leadership in R1, but importantly, translate the leadership that we have on R1, even a fraction of that, I should say, into the R2 products. The R2 price point starts at $45,000, and the average price of new car in the United States is around $49,000. So R2 is like right bull's eye, in the medium market. And for us, that represents the really significant scaling potential for the company. So the guidance we've had on R1 is really just trying to reflect what's happening to the size of this premium market. And is it going to be a little bit more compressed as folks instead of buying a $90,000 car might buy a $50,000 or $60,000 car, or hold off on a purchase entirely until there's a little bit more stability in the system. But importantly, and I'm sure this is being talked about in the conference today is, like overall volume has not dropped. So like the auto industry in terms of volume is doing well. It's just the mix shift has moved into less of these premium vehicles and more into the, call it, $45,000 to $55,000 vehicles.

And I do want to touch on the R2, but I guess maybe just one more point on the guidance that you've given for this year. And as it relates to some of the potential policy, you did guide to about $300 million in the regulatory credits this year. Can you just sort of talk about what are some of the assumptions that were baked into that? Because you saw the Senate pass the Congressional Review Act, and there's at least the -- seems like high potential that carb loses their ability to regulate. Maybe they challenge that, who knows, but I just want to...

They're definitely challenging. Yes. Look, we guided on that. As you said, we talked about $300 million. In that guidance, we contemplate a lot of what's happening. These are -- a lot of this has been -- there's not a lot of surprises happening. The CA waiver getting pulled was expected. It's been talked about for a while. And so that was embedded in our guidance. These are things we expected. We, of course, would like it to not go away, but we are where we are.

So the credits that were sold in the first quarter and over the balance of the year are more at a federal level...

And I think it's important to note, this is -- so while the carb credits and the -- are going to go away and California's ability to require 100% electrification by 2035 is going to go away. The Federal credits, even if they're softened have become on a per credit basis more valuable, because the very rapid retraction from the incumbent OEMs away from electrification. And so what we're seeing is the OEMs are much more aggressively pursuing our credits. And it's really interesting selling credit to other OEMs because you get a really clear picture as to how the other OEMs view electrification. And so like at a working level, the teams are all talking. So I would just say broadly, the significant pullback from OEMs and the refocus on their ICE business, has made the credit's environment more attractive for us. Even despite the fact that you have some credit categories disappearing like EV, BEV credits for California?

That's, I think, a great segue into the R2 or this end of the topic we were talking about earlier because at least from my perspective, it seemed like one thing that has slightly changed. And I'm curious how sort of you view this as a go-to-market strategy is? Right, with the R2 coming in, you mentioned earlier sort of hitting that sweet spot of the market and the price point. But I think you also could have taken the view that it was launching into a U.S. market where everyone was going to have to try to sell more EVs, and you had a more compelling EV offering. If you don't have California, if EPA emissions are sort of not rolled back, but sort of maybe pushed-out and there isn't sort of as much of a need to sort of -- for some of the legacy automakers to move to EV. It seems to me like the R2 may actually be competing more against ICE vehicles than maybe we thought about when the product was initially on that. I don't know if that's your view or not, but I'm curious on your view? And I guess if that is the case, and it seems like the additional challenge here is really to get to convince someone to go electric. And so how do you think about that? How do you think about the go-to-market for R2 in that world?

Yes. It's a great question, and it's important to just recognize where we are in the state of electrification in the United States. So relative to the other large economies, we're by far the slowest to electrify. So like us relative to China, we're half order magnitude behind them. But just in terms of numbers, about 8% of new cars sold in the United States today are electric. Roughly half of those are Tesla. And the fact that, that is the case, the fact that there's this extreme market share concentration with a single company on 2 products, the Model 3 and the Model Y is reflective of a unhealthy lack of consumer choice, which is going to create, I think, this artificial ceiling on how much we can electrify, meaning there's only so many -- the Model Y and Model 3, I'd say, are great products. They're very compelling. But there's only so many customers that want that form factor, that look, that design, that positioning. And so as you start to think about how do you get the other 92% of customers who aren't buying an EV into an EV, we need more choice. In the ICE world at comparable price points, you have 300 different choices. I think it's like 310 right now. And in the EV world, you have 2 good ones from one company. And otherwise, it's a really thin field. And so by the way, that's a sharp contrast to what's happening in China where there's a lot more competition. But in the United States, there's just a lack of competition. I think first and foremost, we need to have choice R2 is, I think, obviously I'm biased, a highly compelling choice at a very similar price point to Model Y. But importantly, very different. Just like the -- if you think of like R1S relative to Model X, they're both 7-passenger SUVs, but they couldn't be more different in almost every way that they think about the decisioning across attributes and product features, application, tonality. Model X is a great cars, it's very, very different than an R1S. And so we have the same dynamic between a Model Y and an R2. And so we really believe, yes, sure, there's going to be some cross-shop with Model Y. But more importantly, our hope is that we pull a lot of people out of ICE vehicles that haven't yet had a product that spoke to them. Maybe they're buying a Toyota R 4, maybe they're buying a Jeep Grand Cherokee, but they're in something that's a more functional SUV in terms of form factor, shape, loading configuration, what have you. And we're seeing that. We see it manifest in just the excitement for the product. We have an R2 in our space in EPAC -- and I was over there actually at a -- like an invite customer event, and it was like the walls were bursting with people. There was so much excitement for the product, and we see it translate to orders. We see it translate to a lot of e-mails that come in every day. And so I say all that, and I would have said the same thing 1 year ago. What's changed between 1 year ago and today, though, is two things which you called out. One is the other vehicles that we both -- that we as a company expected to see come, have taken sort of a back burner in a lot of companies. And then the execution on a lot of those products has been not as strong as what I thought it would be. And the tailwinds that were going to exist in terms of policy driving the continued investment in those products has disappeared. And so I think in some ways, it's a good thing for Rivian, but I think it's a bad thing for the country and it's a bad thing for the world. So I think as i feel for the ones like my kids and wanting a better world for them. I think it's unfortunate that there's going to be less competition and therefore, a slower rate of adoption of EVs in the United States. I think for Rivian is like if you're a shareholder in Rivian, there's no [indiscernible]. It's a good thing. I think it's also probably a good thing for Tesla. But I wish we had more competition. I wish there was a more healthy ecosystem driving us from 8% to 50% over the next 5 to 10 years, but I do think we're going to have a slower rate of adoption because of an underinvestment that's going to happen in the space.

You mentioned R2 orders. I know you gave an update right after the initial R2 launch. Is there any update on traction there? Has that sort of continue to progress upwards?

It's been a lot of excitement, I think, the brand has just translated so well. We thought a lot about like the position of the company and the brand and what we hoped to achieve. And I remember -- feels like yesterday. In like 2017, we would have these brand discussions and product positioning discussions. And a lot of them would end with some version of, if we do things really well, imagine if there's these groups of Rivian customers that get together and self formulate and talk about the product and they're online in chat groups talking about it. Fast forward to today, for the last 2 years in a row, our brand has been by far the #1 rated brand in terms of brand satisfaction on Consumer Reports. But we have one of the most active user bases. As you see in the Reddit forum, you see it. We have Rivian Owner Groups, which are just incredible. And so the level of anticipation around the brand, to now have something that's much more accessible and therefore, much broader aperture, it's been awesome for us as we think about the launch.

There's two more big -- there's a lot more to talk about. There's two more main things I want to make sure we touch on. The first is cost. So you've highlighted some of the cost out as you move from R1, Gen 1 to R1 Gen 2, I think about 20% or so order magnitude is sort of what you've indicated. Curious like if you think there's even more that can be done there on R1, but does it get harder from here to sort of take that cost out? Or what are really the levers on R1? And then moving from R1, let's say, Gen 2 down to R2, obviously, different vehicle, maybe not as premium or as you mentioned, sort of flagship. But I think there's some investor concern or doubt or skepticism, I should say, out there about the ability to sort of take cost down with a lower-priced vehicle and still make it profitable? So just maybe you could talk a little bit about the path on both those?

Yes, you said it. The Gen 1 vehicle with R1, which we launched in 2021. That vehicle, when you think about it, that was sourced in 2018, 2019. So we went out and built all the supplier relationships, signed the supplier contracts for the content that ultimately went into the vehicle, at a time when the auto industry was near peak and at a time where Rivian as a company was very unknown and very unproven in terms of our ability to fully execute the products to launch the product and then have demand for the product. And so when we did that, we had to pay a premium on every part, and we had to swallow it. We didn't know sort of how to -- we had no other way to do it. We needed tires, we needed windshields, we need headlights, we need all the parts. So we took this premium price on all of our content with the expectation that we'd be able to lower that with launch and with demonstrated success. Fast forward to launch, of course, that was now in the middle of COVID. And then as we sort of rounded the corner in COVID, we had this like major supply chain interruption. And rather than being able to negotiate price out as we expected, we were met with suppliers that wanted us to pay a further premium on top of that just to get components, and our lack of scale across the portfolio meant that we were, in some ways, like last to get volume allocation. And for those of who have been following the company, you all saw this, we were like scrapping for every vehicle we could build with lots of missing parts. It was a very -- I couldn't imagine a more painful ramp process than what we went through in 2022 and 2023. Anyway, that led us to deciding to resource a lot of the vehicle. We resource about, well in excess of 50% of the building of materials. As you said, that took around 20% out of our material costs out of our BOM costs. But we did that resourcing activity what we launched last year. Of course, the sourcing didn't happen last year. The sourcing decisions and negotiations happened in 2022 and 2023. But we are doing that in the middle of the supply chain crisis. So while our leverage was demonstrably higher than what it was in, what we did in 2018, 2019. It was still not at the level that would allow us to get truly what I would say, like world-class pricing. Fast forward to R2. R2 is mostly sourced in 2024. The vast majority was completed then and our leverage then was much, much higher. We had a product that was the market share leader in its category. The brand has resonated extremely well. It's really helpful that a lot of the CEOs, a lot of suppliers drive Rivian. So there's just an overall excitement around the brand, the product. We signed a very large joint venture with Volkswagen, which put us from a technical point of view in the decision seat on making a lot of the core decisions in our electrical architecture, not just for our products, but across the whole Volkswagen portfolio. So suddenly negotiating, let's say, electronic components we not only became an important customer because of the Rivian portfolio, we became an extremely important customer because we represented a broader set of volume across this Volkswagen family of products. And so the supplier leverage that we had -- that we have now is like night and day. And I'm giving an anecdotal sort of way to think about this and sourcing R1 initially in 2018, 2019, like the way that worked is like I would have to go to Detroit for the suppliers are based in Detroit, go to the supplier, sit in the lobby, have the meeting start 10 minutes late, go meet with a Director at most of whatever that function was. Following the Director likely telling us some version of this probably won't work, but we're going to help you out here, that kind of a meeting to then when we resourced meeting with VPs, often the VPs coming to meet with us. And now as we sourced R2, the CEOs of the companies, the same exact companies are flying to meet me wherever we are. And so it's just a completely different environment. And so a big part of the cost save is through the supplier negotiations and leverage, and that's not insignificant. We're talking significant double-digit percentage reduction in cost structure. But then in parallel to that, we've just optimized the design in ways that, our engineering organization has matured and grown a lot relative to what we did in R1. And it's every area of the vehicle, it's the big things and the small things, less fasteners, more part consolidation, lots of part elimination. If you want to see a good example of it of like the little hidden details that you don't even notice, you can Google this. Just look at the rear door on an R1S versus the rear door on an R2. On an R1S, there's a dip bar, there's a fixed piece of glass, there's a belt molding and a trim molding. On an R2, it's a single piece of glass. We packaged it also, there's a full drop of a single piece of glass. There's no belt molding. There's no dip bar. There's no fixed glass. And as a customer, you look at -- you don't even notice this. Maybe it looks a little sleeker because there's less parts on the R2, but we took like half the content out of just the door. And that mindset has flowed through every single part of the whole vehicle. And so what we've said is, the R2 material cost is about half the material cost of the Gen 2 R1, which is the improved R1. And then the non-building material costs, so that's all of the assembly costs, our plant logistics, it's well under half of what it is in R1.

For the last topic to close with ADAS, autonomy and AI. And I guess maybe those all sort of build up. So you have sort of the Level 2, Level 2+ features on the vehicles now. I mean I'm wondering if there's any data or insights you could share from your fleet today? Sort of how that's being used, what the reception to it is. You're talking about more Level 3 type features, hands-off eyes-on, I believe, next year, sort of what that entails? And then...

Hands-off, eyes-off...

I misspoke there. And then I guess broadly, how you see that technology progressing for Rivian vehicles and whether you think this can be a differentiating feature? Or what else can come of the data being used and collected from that? And then maybe just -- I know you have the -- you said -- you mentioned there'll be an AI day later this fall. I'm assuming some of this will be spoken about, but sort of maybe a little bit of a sneak peek as to sort of what investors can come to expect as sort of topics for you to address there?

Yes. I mean the first thing just to recognize is, we've had in the world of autonomy, a mega shift in how we approach these systems as an industry. This is like not unique to Rivian, but the way that autonomy was developed prior to 2021, maybe early 2022 was these are -- these were very rules-based systems. So you'd have a perception stack that would identify objects, classify those objects and associate vectors with those objects. All those object identifications plus vectors would be then handed into a planner. And this planner would be something that would be a rules-based environment that would be written by software developers to say, these are the rules of the road. This is how the world works. And that planner would then make a whole series of decisions and projections around the trajectories of those different things in the environment and then, of course, have corresponding intended reactions for the vehicle relative to all those objects in the system. And so a lot of work went into building these planning platforms. And you often maybe hear this call, this is a late fusion process. You'd be taking all this information and you'd be fusing it late in the planner. None of that carries over into what you may heard we call it, what we call is AV 2.0, which is where different than having the information taken from cameras moved linearly into these planners. We now have all the information is taken in very early on is fused into a global view of the world and fed into the build-out of a large parameter foundation model. And that large parameter model is trained "the catchy way that this has been called is end-to-end, so you train it by looking at vehicle drivers, how they behave in these environments across all these different things and you're building this neural net understanding of the world and all the image processing, image and coding is done with transformers. And then obviously, transformer allows it to be ingested into the foundation model, into the neural net. And you have a completely different way of building the systems. So you no longer have programmers who are building rules-based environments. And so the reason it's important to recognize, it's an entirely different topology. It's a complete melt and report in terms of what you would have in terms of technology stack is that, in order to be successful in the long term, our view, there's no possible way, and this is true, in autonomy. This is also true for language. This is true for, I think, almost anything. There's no way for a rules-based programmed environment to beat a neural net, if the neural net has a robust and well-defined data flywheel. And so really, when you think about autonomy, the question is what is the data architecture? What's the data flywheel? What's the training mechanism? And so for us, we re-designed our Gen 2 vehicle contemplating this. And in order to do a robust AI-centric end-to-end approach on training an autonomy platform, first and foremost, you need to have complete control of your perception stack, meaning you can't train an AI model with a perception stack that has a bunch of abstraction layers of third parties. So you can't have cameras that modify information or in any way interpret information or process it and feed into the system because it completely corrupts the ability to do this where you'd be building a model. And so first and foremost, owned perception stack. Second, you need to have a really capable compute platform and inference platform in vehicle that's capable of in real time, running a distilled version of your large model to make decisions around how the vehicle should behave in this environment. Number three, is the vehicle architecture needs to have a really robust way to collect, segment, sequence and choose which data gets saved and then a way to move that data off the vehicle and to do it at scale. And in order to do it at scale, what I mean is do it in a cost-effective way. And so the vehicle architecture should have, of course, a WiFi platform and then the WiFi platform should have a high level of engagement. You need a way to encourage all customers to be on WiFi. So when they come home or park the car in the garage that for [indiscernible] free, you're moving a lot of information off the vehicle into the cloud. And then once the information is off the vehicle, you need a way to train the model. And so of course, the foundation is you need a lot of GPUs, but those GPUs need to be designed. You need to have a team that's capable of running all the experiments. So the right team -- you need the right science team, the right research team. And so we've been building this for the last couple of years. Core to this type of an architecture is, you need data. So you need the car park to be big enough. And so when we launched our Gen 2 vehicle, the features that were in it were built on a really small fleet that was our own fleet. What you're now about to see is, as the fleet has grown, a lot more rapid progression of feature set. And of course, that leads into R2. But what we're planning to go through in our AI Day is to just pull the curtain back. This is one of our largest spend categories in the business. It's something that we -- it's classic if you're building a big project, it's like a housing project, you have to do all the plumbing, all the foundation, with all these things you don't really see. And then suddenly, it seems like overnight, the house is built. And so we've been -- we've got a large team on this. It's one of our biggest teams in the company, huge investment category for us, from a percentage of our R&D. And so AI, sort of coincides with when these features will start to manifest from customer-facing features. So you'll start to see the vehicle get better and better on our Gen 2 fleet. It will make another step forward with our R2 platform. But we want to show how -- we're going to like show how this is happening. So everything from reviewing our hardware stack, what we vertically integrated, how our data flywheel has been architected, what our foundation model architecture looks like, what our approach to GPUs has been, how we're building a large enough cluster to do all this training. And that will be, of course, included with the demo, which is -- it would have to be part of the process. You actually have to see this manifest into a feature.

Great. Well we'll look forward to that. With that, we are out of time. RJ, thanks for joining us. Really appreciate your time.

Thank you. Thanks, everyone.