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

Well, we've provided guidance for the full year that reflects our expectations. We've made a slight tweak to that over the course of this year. But so I'd say guidance wise, we've been pretty consistent. We, of course, are working deal as fast as possible. So there are ways to like go faster and accelerate. We're pushing on those and we know what those levers are. We know what the -- we know with the slowest ramping supplier. We know exactly who they are. We know what the issues are. We have people on site. So we're trying to go as fast as we can because we see the backdrop of so many customers excited to get their cars.

Right. And then the other piece, you talked about adding a second shift by the end of this month, I guess. Is that still on track? Where is that process? Where do you stand in terms of getting that second...

Yes. Second shift is on track. Again, that needs to be coordinated bringing a second shift on to not be busy because you don't have parts, it's also a challenge. So we want to coordinate that really tightly with the ramp-up of the supply chain. But that's been [indiscernible] really nicely. We have a great training program for bringing on new resources both at the team member level, but importantly, at the team leader and group leader level within the plant. And when you think of like plant floor and leadership, this really comes down to the group leaders and the team leaders, this is the engine room, so to speak, of the plan. And so we've invested a lot of time into that function and into those roles into those leadership roles within our operations.

Okay. Great. And then I know the other area that's you're super passionate about is autonomy and the path to autonomous driving, and you guys have communicated a desire to get to point-to-point eye-on driving by the end of the year. Maybe just talk through where you are. Obviously, we're getting closer to the end of the year. What's kind of left to be done to get that rolled out to customer vehicles.

Well, we have -- we're going to announce the specific date, but we're going to have an Autonomy and AI day that we have at the end of this year, similar to what we did last year. And the goal of that is to both provide an update on all the progress that's been made over the course of the year. Of course, since our first Autonomy and AI day, we had some announcements, 1 of which was the partnership with Uber to deploy a robo taxi version of R2 in late 2028. So we want to be able to talk about the progress we're making in the technology and then also you -- think of it as like an anchor point to then roll out some of these new features we're really excited about. And I'd say that on the staircase to Level 4 and the vehicles being capable of delivering robo taxi capabilities or full autonomy. A really important step is point to point. And that's -- and just to make sure everybody is clear what I mean -- what we both mean when we say point to point, this is where you get into the vehicle, you type the address of your intended destination into the screen and say go, and the car does everything from there forward. So you pull the car out of a garage or out of a parking spot, drives the final destination, find a parking spot to final destination. And so that's -- we think that's a big unlock in terms of how people perceive autonomy. It's by no means the final unlock. This is still, call it, early days for what ultimately is going to come, we think in the next few years where cars increasingly, it's an expectation that the car just completely drives itself, meaning it can drive itself empty. It could drive it with you in the back seat. But on the journey to that, this point to point where it's, as you said, supervised. So your hands are off the wheel, but your eyes are still on the road most of the time, when I say most of the time, I mean, you can look down if you distracted for a bit, but you can't take an app. You can't watch a movie. The obvious next step beyond that is an eyes off. So you've heard we probably talk about hands-off eyes off, but the ability to be completely disengaged from the driving. And then, of course, the step beyond that is Level 4, which is you don't even need to be in the car. And so we -- that road map is something we see really like playing out meaningfully over the next 2 to 3 years with 2028 being a really important year for us where we start to deploy the Level 4 version, which is this robotaxi version in a customer application where you actually have it working, driving cars, earning revenue while it's driving.

Yes. And you raised the point about the point-to-point being like the real unlock for customers where it's like a -- there's an added value that comes with that versus just being able to keep it on the same lane or on the highway. That is a very different functionality in technology. And I'm curious, as you've engaged with customers to what extent that is a selling point or are they waiting for that? I'm assuming people who are already in an EV to the extent they're in a Tesla know what it's like to have point to point, but someone coming from an internal combustion vehicle may have never experienced it, they might not know what they're missing at this point. So I'm just curious what the feedback has been in terms of customers wanting to see that on the vehicle before they're ready to make that purchase decision for the R2.

I love this question. I think it's important. And all of you [indiscernible] spend time thinking about this segment. This is a very unique moment in time where in our own -- in our own bubbles, it may seem like everyone is thinking about and knows about autonomy and the nuances of it and the difference between Level 2 and Level 3 and hands-off, eyes-off and hands-off, eyes-on and all these things that I can say quickly and here and we know what I'm talking about. But the number of actual customers that today know about that is actually quite narrow. And so as you said, folks that are in a Tesla because their exposure to FSD, they have more awareness of this. But the vast majority of customers buying new cars in the United States today are not making those purchase decisions around autonomy. That's the way it is today. Now here's what's, I think, the important giant like part that we need to underline in bold faces as autonomy becomes more capable, this becomes a very important purchase criteria because you start to get your time back. And so core thesis that we hold to be true and believe in is that we are going to see an inflection point where this goes from a small subset of the buying public that's early adopters maybe buyers that are buying EVs today to everybody in the United States, everybody in the market is going to want to have the ability to drive to work without having to pay attention to the road. We think that plays out over the next handful of years. And so for that to happen, there's a lot of education, test drives are going to play a key part in this. But we do think it's something that's going to dramatically shift market shares, that's going to dramatically shift the topology of essentially who's providing the technology, who's providing the vehicles. It's going to lead to a very different looking auto industry as we think about the early 2030s and into the mid-2030s. But all that said, right now, the focus is get that technology right. And then, of course, with the recognition that the early doctors know a lot of these details, care a lot about it and then make sure that customers were buying Rivian and buying cars for reasons that are not autonomy, know these features exist because [indiscernible] mouth is going to be powerful.

Yes, absolutely. And I would say would you say that, that is from a pricing standpoint and a willingness to pay for that functionality, where -- what's the balance, right? Because at some point, if it becomes a requirement for this technology to be on a vehicle for it to be deemed even worthy of and being on a consumer's mind, 1 could argue that pricing at some point might have to come down. But I'm just curious, as you think about the next kind of 3 to 5 years, where your pricing is today, what the willingness to pay is and maybe it's a much more complex conversation because you get your time back and you have to factor that in, some productivity gains. How do you think about kind of the pricing algorithm for your technology...

Well, I mean, ultimately, it's going to be market-based. It's -- and so I'll tell you my point of you, but the 2 bookends of how 1 might answer that question is on 1 end, if you -- is the view that autonomy, high levels of autonomy is going to be -- it's a scarce technology. And so there's only a few providers of that. And the other end is it's commoditized. probably not too surprising. I believe and Rivian strategy is built around this belief is that we think it's going to be a fairly scarce resource, and we think it's a resource that's valuable to own. And so we're spending a lot of money and time developing that. I think the number of companies that will likely have it is certainly less than a number of fingers on hand. It's like -- it's a small number of companies that have a legitimate and very plausible path to get to these levels, meaning you have robust data flywheel, access to low-cost inference for running large models in the vehicle, really highly robust perception platform camera heavy, but in our case, camera heavy plus radar and LiDAR, both all 3 of those perception platforms being low cost. And so we think there's a narrow set of companies that fall into that category. And so as a result, we do believe there's going to be persistent pricing power that exists for customers who want to pay for this. And I think, again, this is my point of view and Rivian's point of view. I think every manufacturer to maintain market share as we think about the next 5 to 10 years. We'll need to have software-defined architecture, so the vehicle needs to be software-defined, which today is really just Tesla and Rivian -- when I say that, meaning not a collection of third-party ECUs that run independent OS islands, but rather consolidated set of ECUs that run a common OS. And so today, 2 companies do that outside of China, [indiscernible]. And on top of that, a self-driving platform is capable of delivering very high functions. So ultimately, Level 4 functions. And so I think if you don't have those 2 capabilities, you will lose market share. And so I think every -- that's a belief. So on the back of that belief, I think every manufacturer has 3 choices. First number 1 is -- 3 paths you can take. You can ignore it. You can say I don't believe that those things matter for market share. We're not going to focus on that. you can build it, you can develop it yourself, which requires very capable teams around software, AI, electronics, model building or you can buy it. And there's going to be a fairly, I think, limited number of folks you can buy that from. But what I just described in that framework is exactly the framework that led to our deal with Volkswagen, where Volkswagen decided on the software OS, the first of those 2 big buckets I talked about to say we're going to work with a company that's proven they can do it. I think -- there's lots of opportunities in the software world to do that. In the software platform, I think there's going to be a similar set of big opportunities for autonomy. And as a result, for us, we can monetize this technology platform that we're building either through market share or we can monetize it through providing it to other OEMs or we can do both, which is, of course, you can see thus far, our strategy is to pursue both, both pursue growing market share, growing revenue through vehicle sales, but also recognizing that these technologies can be sold independently to manufacturers that have decided to not build it themselves and have decided to act as opposed to nor these big mega trends.

Yes, absolutely. And I would say or ask, like -- when you think about the robotaxi partnership with Uber, you think about -- you talked a little bit about tech licensing. But if you go to the other direction, you talk about your partnership with Uber, to what extent is that program important for you to achieve or at least expedite your path to eyes off point to point in terms of the data capture you get in terms of the experience and having these vehicles on the road and the driverless on supervised capacity. Is that a requirement to get to eyes off for personal vehicles or just expediting things for you.

Well, it's a great question. It's probably helpful for me to just maybe ground us all to describe how we've approached self-driving. And so -- when you think about self-driving and the technology necessary to deliver on this, all of us recognize that this has been something that's been worked on for a while now since, call it, I don't know, 2012, 2013, when you start to see a lot of conservative efforts around this. And it would be easy but wrong to assume that the work that was happening in 2013 was highly related to the work that would be happening today. And the reason I say it'd be easy but wrong is, of course, the output, the application is the same. It's the vehicle drives itself. The difference is we had a significant shift in how we approach building these systems. And so the way self-driving was approached up until just a couple of years ago, was a vehicle and I've a perception platform on board. So cameras at a minimum, but at a maximum cameras plus radar plus LIDAR. That perception platform would identify and classify objects. So that's a car, that's a bike, that's a tree and then would associate vectors to those objects. So acceleration in velocity, it's all those different objects. And so what would then happen is that set of classified vector associated objects would be handed to a planner. And what the planner was is it was think of it as a large code base that was written by humans then describe driving. So we describe what to do with all these various actors in the system. And the limitations of that were obvious. It was very much specific to the vehicle application. It was very much specific to the geography and the area that it was operating, and it's quite fragile. So if you took that platform put on a different vehicle, it required a lot of work to redescribe how to drive the car. If you took that platform, moved it to a different geography, a lot of work to describe how to behave in that environment. Of course, what happened and we witnessed this in the most hepatical way imaginable with LLM is the introduction of transformer-based encoding, the idea of building large multidimensional neural networks -- and the concept, which is not so common for us to talk about foundation models and this idea that a much more complex multidimensional model can capture a very complex bodies of human knowledge in ways that are not this sort of rules-based approach that we've tried to do in the past. So of course, in the LLM world, you saw -- we saw things like Alexa and Siri in 2017, 2018, which could maybe tell you a joke, maybe tell you the time of day, now advance to LLM, which are like there's absolutely 0 relationship between what was there in 2017, 2018 in LLM. But LLM can give you relationship advice that can write a book report for you. It can diagnose a red spot on your arm. It can do all these things that -- of course, you would never even think to ask Alexa a decade ago. And so I call out the Alexa vs LLM because I think it's illustrative because that's the same dynamic of what existed for autonomy pre-transformer. And so all the work that happened from the 2010 up through early 2020s was what I'll call autonomy 1.0. We're now on autonomy 2.0, where we're not trying to write down in a set of code like how to drive. But we're collecting very large pools of data. And so to do that, you need a big data flywheel. Of course, 2 companies that have effective data machines are Tesla and Rivian. Tesla is, of course, the biggest because it's fleets by far the largest. That -- all that data gets ingested. We use that to pretrain a big model. In our case, we call that LDM large driving model. But there's only 1 model. We don't build an LDM for Level 2+ and an LDM for Level 4. It's 1 big model. And in our case, we're feeding that model with a really optimized set of perception. So we have a set of cameras that's been designed not just to drive the vehicle, but it's designed to learn as fast as possible. So we place them and decide to them to absorb all this information -- we've -- alongside those, we've put those with radars, we have 5 radar in the vehicle. And then soon, we're going to launch a LiDAR. And the LIDAR turns every single vehicle that has 1 that's deployed into part of this really large ground truth data collection fleet for training our foundation model. And anyways, your question was coming back to that. Your question was, does point to point, is that a necessary step or like an accidental step, if you will, to get to Level 4. It's just part of the continuum. And so as we think about it, LDM just needs to keep getting better. And this was the part that's, I think, perhaps the most confusing for customers is if you're in a Level 2+ vehicle, the way that it feels is actually should be nearly identical to a Level 4. The difference is just how confident we are in the system such that will allow you as the operator to cease to be part of the operational control of the vehicle. So Level 2 plus and Level 4, the difference is you don't have to be in the car, you don't have to pay attention. But the car is going to be doing for the vast, vast, vast majority of the time, an identical set of tasks but it's effectively a look at how capable we are in handling the corner cases and then, therefore, managing those in a safe way. And the first step to that is allowing you to take your eyes off the road to say, "Hey, look, the car is very capable such you don't even need to pay attention, so it's no longer supervised." Then the next is you don't even to be in the car because if it does run in -- it's not going to run into an issuer it needs to ask you to retake control. All of that said, one of the things I do think is worth calling out, which gets back to my point on the shift from early adoption to mass adoption, if you're a tech early adopter, if you're one of us, you can live in this world where it's supervised, you have this partial role for a lot of people, that's like confusing. But soon, that will just be like, remember when like we used to have to sit in the car and pay attention while it was driving. That would be like this little brief -- we should all like enjoy this while it's happening because we've got to tell people we used to do this. But this is going to be like a very brief moment, a spec in time before we get to a world where it's just like, of course, the car drives something, kidding me like you guys had to like operate the car. It will seem so foreign to like my kids will look at that and just be blown away that that was the way it was.

Are they going to have a license, do you think -- what's your prediction?

I think they will because I'm influencing them because I like to drive. But I don't think they'll need to in 10 years, I don't think they'll need to. I think it will be -- that's really driving a manual transmission to drive one, but people do it because it's good to do. It's like it's a novelty. It's a novelty, yes. But anyway, long way of saying, point-to-point is a key step on the journey, but it's not the end step. It's just -- it's 1 of the steps. .

That's great. Well, I think we're out of time here. So RJ, I really appreciate you being here. If you have not driven the car yet, go upstairs, take a test drive towards it, but I appreciate it. Thank you.