The New Role of AI in Electric Vehicles
Source: https://www.nvidia.com/en-us/on-demand/session/gtc25-s73711/
RJ Scaringe, Founder and CEO , Rivian
Rishi Dhall, VP, Automotive Business, NVIDIA
The New Role of AI in Electric Vehicles
It's so important that as an industry we create solutions or products that are really exciting to customers. Products that compel them to make the switch from an internal combustion engine vehicle to an electric vehicle. And getting buyers of vehicles exposed to the benefits of a well-developed, highly compelling EV that fits their form factor needs, their price point needs, their storage and cargo capacity needs, is really important. The feeling of acceleration, the feeling of a completely redefined software architecture and software experience in the vehicle and also the way that we approach the vehicle getting better and better over time.
Well the car of today is so different from the car of before and technology is really one of the biggest differentiators. So if you look at the complexity of semiconductors that's available in the car today there's nothing that compares. For many of our customers that will be the most powerful computer that they own will be actually their Rivian. With our vehicles being software-defined, it gives us the opportunity to look at design in different ways that we haven't been able to do before. We have Rivian DNA and we wanted to look and feel like a Rivian and people need to experience it like a Rivian.
We have an architecture where we are able to continuously improve the software over and over and faster with the new generation. When we do the hardware ourselves, we actually have that flexibility. So you buy a vehicle on day one and on day 300, it's a different vehicle, it's a better vehicle.
We can talk about the incredible performance with our new quad motor, which does a quarter mile in ten and a half seconds. We can talk about the massive cost reductions we've achieved in our bill of materials. This part here takes close to 20 components and consolidates it down to one. We can talk about how it's easier to build. This system automatically goes through and marries onto the chassis. It's all automation, it's smoother, it's more seamless. But one of the things we also really want to make sure is clear is just how the program timeline was managed. Not only achieving significant cost improvements, but also laying out the architecture for future products, R2, R3, and beyond.
Adding hundreds of robots, changing some of the flow, the layout, improving the efficiency of how we build the vehicles, which is ultimately going to allow us to run the plant 30% faster. In-house testing in real time, faster troubleshooting so we can reduce the footprint while increasing the output.
We're making thousands and thousands of decisions every day, and the element of alignment that we see across them is a desire to make sure that we're building a company that helps drive towards a better world. What defines a Rivian can really stretch beyond what we think it can be today. We already have a vehicle people love. We're making it more accessible. We're making it more feature-rich, more capable. We have a motivated team working together every day to deliver the best experience to our customers. When we wake up in the morning, this is how we operate as a company. What we've learned in terms of launching a complex product, turning on a supply chain, ramping the supply chain, is all coming together as we now think about the next phase of growth with how far we see the ability to push our company and our products beyond what we've seen with R1.
Rishi: Rishi, over to you.
Rishi: Hi guys, welcome to GTC. We're so proud and we welcome RJ. Co-founder and founder and CEO of Rivian. RJ, thank you so much for being here with us.
RJ: Thanks for having me. Yeah, this has been exciting so far. We had a great keynote by Jensen and, you know, AI is everywhere. This used to be a GPU show with a little bit of AI. Now it's 99.9% AI and we're so excited that you've integrated so much of AI within your company and we're going to have a little bit of conversation around that.
So, Rivian just posted its first positive gross profit and continues to scale production. That's a massive milestone, by the way, so congratulations on that. What would you say has been the hardest part of getting here? And what are some things you underestimated when you started the company?
RJ: Oh boy, a long list there. I think in building something like a car company and something that's both complex as a business but incredibly complex as a product, you know, in the beginning on day one you have none of those ingredients. So you need many billions of dollars in capital, you need... You need hundreds of suppliers providing thousands of parts. You need a team of thousands of engineers, a brand that people care about, a product that connects with consumers in a powerful way. And so to start, you have zero. And so it's really an interesting exercise of you almost have to will some of the ingredients to start to come together because you don't have capital, so you can't develop technology. You don't have technology, so it's hard to raise capital. But the first few years, you know, we sort of had to just initiate some early momentum around the idea of a product. We went through a lot of really big pivots in terms of what the brand was going to become, the products we were going to develop would ultimately become. But there's a few elements that from, I started the company in 2009, that from 2009 through to today have been really core and one of those was like a deep belief that the products had to have meaning beyond the technology, meaning they needed to emotionally connect with consumers.
But we felt that to do that, it was really important we vertically integrated in a number of areas. And for us, one of the most important areas was the electronics and the software that go into the vehicle. And so you can imagine, you know, in, I don't know, 10 years ago, I'm talking to our then very small investor base and saying, we're going to go develop all of our own computers, we're going to build our own software system, you know, so our own real-time operating system, our own operating system for infotainment platform, and the investor's like, you know, you really want to do all this? You know, that's a lot of work. It ultimately really proved to be the right strategy we've the vehicles and the brand that we've created around them has generated so much excitement and we have the last two years during the highest level of customer satisfaction there's a few independent parties that rate this. But we've also been seeing a great deal of validation just in terms of that software stack being applied to other brands. And so in the fall last year, we signed a $5.8 billion software licensing deal as part of a joint venture with Volkswagen Group. And so it's so cool to see all these things that we've been working on now for so long start to be deployed, not just in Rivian products, but also in different brands, different form factors, different price points across the world through this big relationship we have with Volkswagen Group.
Rishi: So, we'll talk about Volkswagen in a minute, but reflecting on your journey from MIT, 2009 you started your company, how have advancements in AI and technology influenced your vision of the company? Like, back in 2009, I wouldn't have imagined, you know, generative AI being such an important part of your industry. Today, there's nothing that you can do without it, so...
RJ: Yeah, I mean, in the beginning, the focus was really around innovation around the vehicle, the propulsion, of course, a huge focus on sustainability. And what's happened in the last, starting maybe 10 years ago, but really ramping up over the last few years, is just an intense focus on how AI integrates across the entirety of the business. So in a very obvious way, it integrates into the vehicle where the vehicle increasingly drives itself. With our Gen 2 vehicles, which is a relaunch of what we launched initially, our R1T and R1S, we have much higher levels of compute, incredible perception stack. Those vehicles have a hands-free feature for when you're on the highway, you can take your hands off the wheel, the vehicle drives itself. And about 20% of the miles driven in our Gen 2 fleet, the vehicle is driving itself. And so in the next couple of years, we think that's going to grow from 20% to 60, 70%. So it's happening really quickly. And we can talk about just how the approach to developing self-driving has evolved so much in the last few years. But beyond that, it's in every part of our business. So if you think about us building a very large... Customer-facing, direct-to-consumer brand, we have to build service infrastructure, which means there's a large diagnostics team, there's a customer service team, we have lots of service technicians, we have hundreds of locations, Rivian locations that do different tasks, whether it be charging, service, sales, distribution. And so the beauty is we're actually designing that architecture for all this infrastructure through the lens of a world where AI is a really key part of it. And so we don't have to unbuild decades of infrastructure to rebuild it with AI. We're actually really designing most of it from the beginning, fully contemplating what it can become in a world where you have much better access to information. Things that you would think about doing manually, you can now automate entirely, and so we're really excited about what that unlocks for us as a business.
Rishi: That's awesome, yeah. And talking about AI, when you look at China and the EV innovation that's happening there, at lightning speed by the way, you know, I saw some Chinese colleagues of mine over here who were visiting for the conference. Tons of new models, advanced battery technology, a lot of AI, and such lower prices. So, in the West, we have far fewer choices, higher costs and slow progress. Why is China pulling ahead and what can Rivian and other companies in the West learn from them?
RJ: Yeah, there's a few questions embedded in that, so I think the first thing you said, which I'll call out, in the United States today, when you think about electrification, where 8% of new vehicle sales are electric, so said differently, 92% of the cars sold in the United States are still combustion powered. And a big driver, we can spend hours talking about.
Rishi: By the way, 45% in China this year, or last year.
RJ: So as a reference, China's at five times where we are in the US, five, six times. And when you look at what's causing that. You can debate the causality. I think a really significant part of it is there's not a lot of product choice here in the United States. And so if you think about where most vehicle sales are occurring, it's under $50,000. And in that price range, there's to date a handful, that's too generous, one or two great, highly compelling choices. Of course, one of those being Tesla. With the Model 3 and the Model Y. And that's evidenced by the extreme market share concentration you see with Tesla. And while the products are great, if you compare that with the internal combustion world, you have hundreds of choices. And so you have different form factors, different price points, different brands, different feature combinations. And in order for us to see significant EV penetration in the United States, we're going to need a lot more choice.
And so we launched our brand with the R1 product line, which you see on the screen there, and that's a premium flagship product, the average selling price of it is close to $90,000. The SUV is the best-selling premium electric SUV in the country, but it's just a relatively small market because there's not a huge number of consumers that are going to spend more than $70,000 on a vehicle. And so our next product, which we call R2, opens that up dramatically with a base price of $45,000. It takes the magic of what is a Rivian at that higher price and puts it into a slightly smaller package, but just it's the coolest vehicle. So we're really excited about that, but we need lots of those. We need our R2 to be successful and we need 10, 15, 20 other options in order to see penetration of electrification really grow here in the US.
But the other part of your question was what was around what's happening in China. So China's stark contrast to the United States where you see a lot of different choices, a lot of different brands, a tremendous amount of investment. And what's what's really unique about what's happened in China is I think we've often focused on the cost advantage that exists in China. And I think that that's of note today, but over time that will fade as manufacturing and the industrialization of building vehicles becomes more and more automated. The relative difference in labor costs will start to fade. And what we're left with is how do the products compare in terms of features and content, which is technology, which is thoughtfulness around how the brand comes together. And that's what's actually really interesting about what's happened in China. In the Western world, there's two companies, us and Tesla, that have redefined the network architecture, built vertically integrated compute stacks, vertically integrated software. Whereas in China, there's a lot of companies that have gone about doing it from a clean sheet perspective.
And the reason that's so important, ironically, actually goes back to fuel injection systems. So prior to 1960 cars didn't have computers in them so they were completely analog and so you know the the engine was operated with a carburetor and everything was very mechanical and so the first computer that made its way into the car was actually a computer to run, again ironically, the fuel injection system and you wind the clock back 65, 70 years, at that time that wasn't viewed as a core strength so that technology was handed to a supplier, you know, so companies like Bosch really grew dramatically with this. And over a period of many decades, more and more systems started to have little computers associated with them. So in the auto industry, these are called ECUs. You know, in these little, you know, electronic control units started to pop up. So you got a set of powered seats, the powered seats came with a little computer. You had a fuel injection system, had a little computer. The HVAC system had a little computer. And in a very organic and non-designed way, we ended up with these horrifically complex and archaic ways of thinking about a network design. So you have a network that's comprised of maybe a hundred of these small little computers, all of which are running different sets of software that are all little islands of software, all developed by tier one, tier two, and sometimes tier three suppliers. And so it's not at all surprising that if you wanted to do an update, Central computer.
RJ: Yeah, so in the most extreme version, you have one computer that does everything for practical reasons and wiring harnessing reasons. You end up with, in our case, we have three what we call zonal computers, two in the one of the front right, one in the front left and one in the back. And so you massively shrink down the number of computers that are in the vehicle. But to do that, you have to design the computers, you have to build the software stack, and you're doing a lot of the content that used to be owned by suppliers. And so in the West, there's two companies that have architectures that are doing that, us and Tesla. In China, there are many. And I say this all the time, but in order to be truly competitive in AI, and to really leverage AI beyond a very narrow set of features, like let's say self-driving, for it to rethink the whole experience, you have to have the plumbing, right? You have to have a network architecture that promotes a common software architecture. You need to be able to issue updates across the whole platform, and this is really hard when you have a system that's so embedded and built around tier one, tier two suppliers. So everything I just described was the reason we did this very large deal with Volkswagen, it was to take this architecture we developed and deploy it within their business.
Rishi: Great answer. So talking about Volkswagen and the deal, as we know from China and Tesla and you guys, that AI and EVs are reshaping the entire auto industry. But there's a problem with the legacy automakers that are struggling. Why do you think traditional car companies still don't get about this transition? What do they not get about this transition? And why do they have to partner with companies like you instead of doing it themselves?
RJ: Yeah, I mean, um... I think it's a very practical, real challenge. If you imagine yourself inside of a large car company that's maybe 50, 60, maybe 100 years old, there's decades and decades of decisions and organizational structure that have been built around a core set of skills, largely mechanical, so designing engines and designing bodies, that does not matter in the future state. They're like engine design skill is not something that translates to, you know, designing compute platforms or designing software. And so these very large businesses have to rethink their own makeup to say, how do we become much more of a software company, and much more an electronics company? And it's just that the practicalities of trying to convert those large businesses to become that is real and what exacerbates this challenge is we have a very large, the auto industry is a very large supply base that has increasingly had a very messy approach to software where you have lots of suppliers of lots of content that want to hold on to that content and that content in terms of these little ECUs. Usually isn't even programmed by the supplier themselves, which is sort of wild. So there's like layers of abstraction. So the OEMs write requirements they hand to the suppliers. The suppliers take those requirements, reinterpret them, hand them to their suppliers and their suppliers usually write the software. Tier 1, tier 2, tier 3 just keeps going. So it's really hard to break the system. And I think being a new entrant, being a new company, provides a clean sheet of clean canvas to build this totally differently. You would never architect a business the way that the auto industry has been built from a software and electronics point of view.
Now, the thing that makes auto so hard, and I can certainly vouch for this, is we also have to build large mechanical systems, so we're still building, we have to stamp or cast or... Fabricate bodies, put them together with high dimensional accuracy, manage really complex supply chains. And these are the things traditional car companies have done for a long time and are quite good at. And so what we're witnessing is traditional car companies trying to figure out how to do technology. And companies like like us, very strong in technology, trying to figure out how to build cars at scale.
Rishi: That's right. Yeah, and you know, we're hoping to figure that out really quickly. That's perfectly right for my next question that I was going to ask you that, you know, scaling production is obviously, you attested that how difficult it is. We've often heard a famous CEO state, you know, "software is easy, hardware is hard." So, how do you prevent companies like yourselves failing into the good tech, bad execution trap that's hurt so many EV companies?
RJ: The first thing in building a car company is there needs to be, you have to be intellectually honest with how hard it is, and it's not something you can do today, maybe 15 years ago is different, but today to do this you can't do it for a small amount of money, meaning you're now competing against both incumbents that are trying to electrify and doing some things I just spoke to, trying to develop technology or rethink their technology stacks, as well as Tesla. And as you noted, while tariffs are maybe protecting in the moment, on an international basis, you have a thriving technology stack and EV community within China. And so this is a multi-billion dollar exercise and requires really strong and robust teams of people to come together to work towards it.
And in our case, with the benefit of hindsight, we launched in end of 2021. The first product we launched was that vehicle on the screen there, the R1TR truck. About a month later, we launched the SUV. And about a month after that, after that, we launched a commercial van that we developed with Amazon. And that was a lot, that launching any product, particularly something as complex as this is your first thing, is a really complex endeavor. Doing three at the same time is really, really complex. And then layer on top of that, we were doing it in the middle of COVID. And so, uh, the supply chain crisis that we all felt where it was hard to buy drywall in 2022 we were feeling across you know each vehicle has around 30,000 discrete components and any single one of those components can stop production so it can be a single fastener that stops the whole line. And I think one of the things that often gets talked about in ramping up production of something like this, something as complex as this, is what happens to production output when you're short one item. And that's obvious. You don't make as many cars. That hurts you on revenue, it throws off your cost structure. The thing that's less obvious is most of the rest of the parts still come. So, we had shortages on about 2% of the vehicle.
Rishi: Which is great. You said we had 98%.
RJ: No, the problem is we had 98%. So we had warehouses filling up of parts because of supply relationships. And we had to deal with that for our first time launching, coming out. And so I was just saying to someone, our, you know, our supply chain ramp team was forged in this like crucible of pain in 2022 and 2023 so as we're now thinking about launching our next product it's a whole new plant we're building this R2 product it's a very different supply chain we have so much more leverage we're so much more mature as a company. But the thing that's, I think, the hardest about ramping this type of a business is actually the supply chain. We've got so many thousands of parts coming from hundreds of different companies that need to be coordinated perfectly. And once the system starts going, then it sort of has some internal inertia and momentum. But in the beginning, you know, you get one phone call from one supplier and says, hey, we've got an issue. One of our presses is down and you're like in your head calculating how many millions of dollars that cost you an hour. And it's, you know, it's a challenge.
Rishi: Not easy, not easy. All right, so this one, wanna talk a little bit about our relationship together. So at NVIDIA, we believe in the three computer problem. We believe that you need training systems, so GPUs for training. Everything needs to be simulated, digital twin of your factory, digital twin of your car, simulation environment. And the third computer is your edge computer, which goes inside your car, or your factory, if that's what you're doing. So we worked closely the last few years on establishing our relationship on all three levels. So you're using GPUs for compute, we talked about Omniverse as a digital twin platform that you're evaluating, and obviously your cars have NVIDIA or inside them. So how's that journey been and talk a little bit about how having the same architecture across the three computers which is all CUDA, all GPU has benefited your development environment.
RJ: Well, I think it's worth also just calling out how much has changed in the last few years. So if we look at the way self-driving, I'll talk about self-driving first, because I think that maybe is the most relevant for the conversation. The way self-driving was developed for, call it the first 10 years of self-driving being a topic, was it was very rules-based. So you'd have a perception stack, so cameras, at a minimum cameras, cameras plus perhaps radar plus perhaps lidar in our case camera plus radar. That sees the world and classifies all the objects that are seen. So, you know, that's a car, that's a bike, that's a motorcycle. And then along with those classifications we attach a vector. So it's the velocity and acceleration of each of those objects. And all those objects are then fed into a planner, and the planner then makes a set of rules-based decisions around what to do. And then the planner then communicates or hands off to execute the controls to essentially tell the vehicle, steer, accelerate, brake, what have you. And that was the way these systems were largely built, including our own, our Gen 1 vehicle used, architecture was a lot like that.
I guess, what, three, four years ago, we saw this change coming that we're now experiencing, which is an approach of using a much more end-to-end philosophy, where you spend less time in these sequential steps worrying about exactly what each object is classified by, and rather train, build a large data flywheel, and you train a large offline model that you can then distill down into an online model, you know, the inference platform that's running in the vehicle. And the speed at which we can make progress is remarkable and very different than what was built before and of course now is very data-hungry, but data-hungry in terms of the fleet needing to be able to provide useful and valuable data back, and then our ability to consume that data. So as evidenced by NVIDIA's stock price, you were right on both sides of that. So you have both, of course, the GPUs that are necessary for this type of an AI approach. And this approach we've seen play out to an extreme degree with LLMs. And the unique thing about LLMs relative to vehicles is there's a handful of big foundation models being built, but they're all processing a very similar data set. So they have the benefit of being able to crunch the internet. In self-driving, we don't have an internet of usable data. And so the data is something that we generate. It, of course, is noisy. We have to find ways to thoughtfully pull out the useful pieces of data. And then we need to run that through. Our training flywheel and so to your point we work with you on GPUs for offline training and then in the vehicle we designed a much higher capability platform with our Gen 2 which has 240 tops which is a lot more than what we launched with on the Oren platform to do all that in-vehicle processing all the inference so that the real-time decisioning that's happening is from that platform. And the roadmap, of course, is everything grows. So we're going to have more powerful platforms that exist in our future products. And similarly, we'll have more and more GPUs offline training our large parameter model.
And the really unique thing here is as we start to think about building these foundation models for the physical world, of course, our immediate application that we're all drawn to is the vehicle drive itself, increasingly well in a broader and broader set of situations. That physical world foundation model will have applications elsewhere. So you can think of embodied AI and how that's going to manifest in our lives, which we haven't really seen yet. Even when we see robotics like that video we showed at the start, which had a bunch of robotics. Those are robots, yes, but there's no AI in them. They're moving to prescribed, predefined paths. And what's really exciting is to start to think about how the ability to have reasoning and decisioning built into these platforms based upon a global understanding of the world, how that shifts everything in the physical world. All forms of robotics, robotics introduction, being introduced in new places. And really, a self-driving vehicle is just a robotic application with a really complex environment. Lots of things can happen in an environment. But with extremely simple actuation, meaning you can speed up or slow down, so you have longitudinal control and you have steering. So there's a really narrow set of actuators, but a very, very complex environment.
Rishi: Yeah, you've built a great team, and it's a pleasure to work with James and NVIDIA, who are doing an amazing job inside of Rivian and creating some amazing products that we can't wait to get our hands on. So talking about robots, humanoids, something that you guys have been thinking about? Because Jensen often says that the best people to build robots are the car manufacturers, because they already know how to build, you know, machines, they're building a different kind of robot for the road, just a different kind of robot. So is something in your line of sight?
RJ: Well, we, I think as you look at robotics, we there's many many forms of robots and humanoid is one and it's it's it's of course one that can be easily applied to lots of different things because the world has so much been designed around human form factor but in particular as you start to think about industrial applications, we think there's going to be a range of robots that have a much higher level of intelligence of what versus what we see today, but that have form factors that are conducive and capabilities that are conducive to more efficient operation within a plant. And you know things as simple as being able to extend in z you know get taller to pick things out or move rapidly with wheels so we we do think that the role of robotics and manufacturing is going to be extraordinarily, uh you know it's going to be an extraordinarily disruptive technology in terms in terms of leveling the cost structure to produce goods. You know, in an environment where we're seeing increased tariffs or the potential of increased tariffs, I think being able to compete in lower cost labor markets by using increased automation is going to become increasingly important.
Rishi: Yeah, makes sense.
RJ: But to say, I mean, there's going to be, just like in the biological world, there's a diverse set of form factors that have evolved, you know, humans being one of those form factors, but like a cheetah can run way faster than I can. So we're not the optimal form factor for every task. And I think in the robotics world, you're going to see a pretty diverse set of evolved form factors that are optimized for different use cases.
Rishi: I'll take that as a yes for humanoids, for RJ. All right, so this is our last section, RJ. Let's do something fun. We'll do a rapid fire. So you have to answer in five seconds or less. You don't have to think too much. And then if we have time left, we can take maybe a couple of questions from the audience. So AI's biggest impact on EV in five years.
RJ: Oh, self-driving.
Rishi: He's the right audience. Okay, this one. Will AI ever make self-driving or driving boring? Five words.
RJ: I don't think so. I think it takes the burden of it gives you your time back. And I said it before, 20% of our miles today are driven autonomously, which is remarkable, because it's a today we just have a feature that's available on the highway. But as we grow our ODD, our operational design domain, to be turn by turn, this number is going to very quickly be 75%. Or higher, depending on the customer, but the ability to get time back is really powerful.
Rishi: Yeah, I always say this to my partners when I talk to them about self-driving, that it's the ultimate luxury. Like, yeah, you can have great seats and everything, but the ability for you to drive when you want to, versus when you have to, that's what luxury is all about, so I agree with you. What's one AI-powered feature that you wish Rivian vehicles had right now, but isn't possible yet?
RJ: Sorry, I'm bragging. Well, he's almost there. I don't know if... I guess you used the word yet when you said it's impossible yet. I don't think anything is impossible. I think it's all just a matter of time. We didn't talk about it, but it is worth noting. We talked about increasing compute levels, increasing inference from a self-driving point of view. The other thing that's happening is we have... These are incredibly powerful inference platforms that are going to be part of the in-vehicle experience and what that will really change it will become so natural to just interact with the vehicle in a normal way as if it's a as if it's you know somebody sitting in the car next to you so instead of having to be able to type in the address you know and figure out like the first half of the address the vehicle figures out the rest which is already too much to just say I'm hungry. The vehicle respond back and say what do you feel like you say i don't know yeah and you can reason your way to a solution in the same way that we've grown accustomed to interacting with you know GPT and platforms like that the the vehicle will have that so deeply integrated that the ways of interacting with the car today, which seems so normal in five years, are going to feel archaic, and I think that's true for a lot of technologies that we interact with on a daily basis, where the ability for the technology to just know what you're trying to do. I had a review earlier today with some of the folks on our team, and finding parking is a pain. And to a human you can explain it so easily. I'm going to this event. Parking is going to be a pain. Help me figure this out. But I don't have a way to put that into the screen right now. But we're working on that. That will come very soon.
Rishi: Great answer. Outside of self-driving, if you could steal one piece of technology from another automaker, what would it be?
RJ: Hmm, let's see. I think the thing that we look forward to having the most is the scale and the cost structure that comes with scale. And that, of course, will be born out of launching R2, which is a much lower priced product, which will give us a lot more volume. But supplier leverage is a really nice thing. And we are on the receiving side of it. You know, in 2022, 2023, I remember we thought we'd be seeing suppliers take cost out as our product was successful with, you know, we're like best selling premium SUV. We're like, oh, this is great. They're going to work with us, take cost out. And instead, in the middle of the supply chain crisis, they said, if you want parts, you need to pay us even more. And, you know, of course, our size and our leverage sort of was what what is the bigger challenge? Making EVs more affordable or making them truly autonomous?
RJ: They're independent.
Rishi: I know, I know. I'm just thinking which one is more like a challenge for your team to achieve.
RJ: They're both happening. So I think an interesting question is, do autonomous features translate across all price points? And so for us today, moving from a high-priced flagship product to something that's going to be starting at $45,000, so still at $45,000, that's much cheaper. The average selling price of a new car in the United States is around $49,000, so it's right in the meat of the middle of the market. But to get to like a $20,000, $25,000 EV. There's a lot of cost out still and that creates pressure on every system everything from door handles to tires to perception stack to compute and so we've we're focused on delivering that first in a $45,000 car but the next thing you know is we have R2, R3 and then like R4 and R5. Incredibly creative naming we have, but in those future products, we are thinking about how do we get to a product that's in that $20,000-ish price range, $20,000 to $30,000 price range. But that's not something we're even working on today. We're focused on R2 and R3 as it is today.
Rishi: So when you're talking about these different car lines, what will be the most important thing for you for the next five years? More AI in your cars and factories or better batteries or something else? What is that one thing that you want to be able to achieve?
RJ: I think it's a sort of a false binary, you need to have both, we need to be exceptional in AI, I think at all price points, and we need to have products in order to drive 100% electrification over the next, you know, I hope five to 10 years, we need a product that exists at price points, you know, well under $45,000. I think the other thing that's important to note here is we need choice, and I think customers are going to need not just a Rivian or a Tesla to choose from, but we're going to need other companies with really compelling products and great technology in order to fill out enough demand to support 80 to 90 million vehicles being sold globally every year.
Rishi: Any questions in the audience?
Audience Member: Yeah, just one question about, you talked about the architecture you created, the SDV architecture you created, and you did a deal with Volkswagen as well, right? Do you see that as a viable business model to license that operating system and architecture to other OEMs? And the reason I ask this is because everybody's trying to build the same thing, more or less. You know, and replicating the same aspects of things, we look at the architectures underneath. So that would be the question.
RJ: It's a great question. You know, to start, we're, we're focused on our products in the Volkswagen Group products. There's absolutely opportunity with other manufacturers. You know, an interesting thought experiment, which I, it's sort of a riddle that I don't, in fact, myself know how to answer is I would say emphatically and with deep conviction that I can't imagine a car company existing 10 years from now with the archaic network architectures that exist today. Meaning, a platform that doesn't really facilitate deep over-the-air updates, it doesn't facilitate deep integration with AI for which you're coordinating with, you know, maybe 75 to 100 different supplier ECUs to do any sort of deep software changes to the vehicle i just can't i can't envision that still being alive in 10 years but i also don't know how we're going to get to a place where all the vehicles sold don't have that because these are skill sets that are not generally present within traditional car companies the tiered supply base so the big tier ones that make those ECUs have like a business school case study in a conflict of interest where they make a lot of money selling a lot of ECUs, these little individual boxes, and the idea of selling one bigger ECU is highly destructive to the revenue. I mean, to make the point, we launched our vehicle, our Gen 1 vehicle, had 17 in-house ECUs, and we did the ECUs in-house with the full understanding that we were going to consolidate them over time. And that was still maybe a factor of three or four less than a traditional vehicle. We then got down to seven ECUs and it saves thousands of dollars per car. Now for us, we love that. That's cost savings. If you're on the revenue side of that, if you're a supplier, it's a lot of business loss. It's a much more efficient architecture. So I think the suppliers have a real deep, true conflict of interest that I don't know how that's navigated in. So I look at it and I say, yeah, I do think there's growth opportunity for us, but it's hard to predict how this can play out.
Rishi: Okay, I think last question guys.
Audience Member: I guess I got it, maybe two or three together you had the SCV lab in Mountain View and up in Redmond, where I am from, is that where the autonomy is being developed and will that licensing that you've done with Volkswagen apply to the whole Volkswagen group including Audi and Porsche and everybody and then last I have a grandson will be 16 in 2028 can you depend on a Rivian AV by then?
RJ: So, yeah, so on the Volkswagen side, yeah, so it's with the group, so it's, which, like, I grew up a huge car enthusiast, and specifically a Porsche enthusiast, so it's really exciting to now, my dad, when we announced the deal, my dad calls me and he says, well, now you'll be able to buy a Porsche, they'll have Rivian technology. So I was like, yeah, exactly. Yeah, so it'd be Porsche, Audi, Volkswagen, Skoda, there's a whole host of brands. And to answer the question for your 16-year-old, yeah, he should, we should definitely be there by 2028.
Rishi: Well, this has been an amazing conversation, RJ. I have to say you've made EV look cool, and somehow you've convinced people all around the world that electric truck is the ultimate adventure vehicle. So it's a pleasure to share the stage with you. Thank you so much.