Beyond EVs: Rivian founder & CEO RJ Scaringe on AI, robotics and the next big bet

27:15Of course, they're not the same people. They're not on the same teams. You want that team to be connected to the team that's doing the drive unit to be doing the software. And so you create frameworks for how you make decisions. And the success of any highly complex product, like a car, I think is ultimately a measure of how coordinated the teams are so that the decisioning feels consistent. Um, so that it feels like, like a very small group of people did it. That's what you want it to present as boy. It looks like 10 people designed this whole thing when in fact maybe 5 ,000 people worked on it.

27:47When you think about the number of decisions and the number of engineering trade-offs and, and, uh, sort of analysis are necessary to complete the product. There's all types of development things that can go wrong. And so that's why we go through multiple iterations of prototypes and testing to build hundreds of cars on the journey to the first saleable unit. That's largely controllable by us. Meaning we, we have visibility to the health and status across. Thousands of different metrics on developing the product, but then there's all the suppliers that make components. So we make a lot of components ourselves, but take, let's say a headlight.

28:26We designed the headlight. We designed the lighting in the headlight. We do all the studies of how it's going to look, how it's going to behave. the control systems for it because you're now using headlights for intelligent. But a third party, a supplier makes it. And so you think of, oh, there's a tier one supplier, but that supplier has suppliers. It has tier two suppliers. And that tier two supplier has suppliers. And those are tier threes. And those tier three suppliers have suppliers. So when you look at the number of companies that are ultimately involved in making like an R2, it's tens of thousands of companies.

28:57And we have maybe four or 500, about 450 direct companies that we work with, like that are our tier one suppliers. They all have anywhere from a handful to maybe 10 plus suppliers that feed them. Those have a handful to 10 plus. So just like really quickly becomes a huge number of companies involved. And in stable supply chain environments, that's all right. So you don't have to spend time getting to know the mine that your raw material is coming out of. But in environments where the supply chain has lots of disruptions, suddenly you have to think about who's my tier four. And the tier four feeds the tier three, the tier three, the tier one, the feeds us.

29:42We experienced that first time when we launched R1. there were so many disruptions to the supply chain with COVID that we had to get very skilled at understanding risk across all of our different tiers of supply. It's quite complex. And so the biggest risk with any program of this type is that you can't predict and you can't actually measure everything happening at every one of your suppliers. And this falls into like big geopolitical challenges. So take a battery, our batteries have nickel in them. 93 % of the the world's nickel comes from one country, it comes from Indonesia. Trade policy with Indonesia could have a dramatic impact on our ability to get nickel.

30:23And that's just one material in a battery. So it's just extrapolate that out. There's a very large team that tracks these things. And we had to learn all those skills. Those are skills you don't, when you're starting a company, you think I'm going to design an engineer. You don't think I'm going to build a team where we have mining expertise and we have shipping logistics expertise. and we have people that are tracking shipping patterns in the ocean and seeing what's being disrupted by conflicts. I mean, there's the amount of complexity to manage all that's quite high. And now we have all those capabilities we've built in, but I'd say that's the biggest risk is just supply chain disruption.

31:00Yeah, I can imagine that this is where you spend a lot of time. And it almost sounds like a big command center, like lots of different experts looking out and keeping tabs on whatever it is they're responsible for, because the complexity is enormous. Yeah, I think, you know, globally, we often talk, I think it's sort of easier to think about, if you think about like supply chains as like this coffee cup, they're quite simple. Like there's maybe just probably one company that does everything on this coffee cup, maybe two or three if you look at the raw materials. But developing complex products, the supply chains are not just dealing with coffee cups, it's really complex supply chains.

31:34And so they're just very multifaceted and they do require you to be like awake at all hours kind of thing. Well, you know, it's funny that you talk about the geopolitics, because I think you can't really have a conversation about business in this day and age without discussing the political climate currently. You know, just briefly, and I'm sure you've been asked this a lot, obviously with EVs, it's an interesting time to be in the space, let's say, right? Because we saw kind of the rollback of the federal EV tax credits. And to your point, there's a lot of going on with trade tensions and whatnot.

32:04How do you think about that? And has it changed the landscape for the industry fundamentally, do you think? Or do you think that Rubian is really just writing this out right now? It's short term. Who knows what could happen? That's a good question. I don't think, I think of it long term, like in the fullness of time. All the policy changes aren't going to change the end state, which is the world's electrifying. Nearly every car, new car produced in the not too distant future. We could debate whether that's five years, 10 years, 20 years away, but certainly in my lifetime, every car produced on the planet, new car will be electric.

32:45And so then the question is like, what's the journey to get there? And if we believe that's what the end state will look like, what's the rate at which it gets there? And I think changes in policy, the biggest impact I've seen is just how much it's affecting other companies. And so the number of large existing manufacturers that have pulled back pretty rapidly from electrification, in many ways, it underscores or underlines the importance of Rivian and a small number of other companies that remain focused on electrification creating choice. So if you're looking at it from a competitive landscape point of view, there's going to be a lot less competition.

33:27If you look at it from a health of the auto industry or from my kids' kids' kids' perspective in terms of the health of our planet or the health of the Western auto industry, I think it's really, like, it's actually dangerous. And we need more companies like Rivian or more companies like Tesla that are building electric vehicles, highly compelling, highly desirable. We hope that the example we set with R2 and we hope that its success will stir up more competition to have more people coming into the space. And interesting you point out, you know, I understand you often say the reason you started the company is it's for our kids, kids, kids.

34:05It's the age old question. But, you know, how did becoming a public company chief change the way that you lead, RJ, especially when it comes to balancing Wall Street's short term expectations with delivering on that long term vision? I mean, this is something I feel like people constantly battle. I think many founders, when they start a business, I'm in this category, where your timescale that you're thinking about extends way beyond a quarter and even way beyond the fiscal year. And so you have a different mental model for making tradeoffs. That's certainly true for me. A friend of mine once gave me the advice of be comfortable being misunderstood for a while.

34:50And so what he was speaking to is we're investing heavily into technology. We're very vertically integrated. We design all of our electronics in-house. We build all of our software in-house. We build all of our power electronics and high-voltage systems in-house. And to do that, we have a very large engineering team, and therefore have a very large OPEX. We're an R &D budget for it. And the reason we're doing that is because we're planning to become a very large company, meaning we expect to build many millions of units a year, but we're not yet. But in order to become a company that's building many millions of units a year, you need to have those technologies and to be world-class in those areas.

35:30And so as a result, our business, like the revenue isn't keeping up with our OPEX and it will soon. It's not like this is an accident. This is highly planned. We like very intentionally said, we're going to raise a lot of money. We're going to deploy a lot of capital building all the infrastructure to have this highly vertically integrated business from a technology point of view and from a customer point of view, meaning we own sales, service, distribution, all these go-to-market functions, which in the beginning are these like structural cost disadvantages because you have really high fixed costs and you don't have a lot of volume to cover it.

36:06But as soon as we get to a certain level of volume, it all starts to really make sense. And so that for us is all embodied in R2. As a company, they're managing public markets. There are times in the public markets where there's like a lot of growth investment. And there are times in public markets where there's much more focus on short-term profitability. Rivian went public at a time when we were coming off the back of a lot of focus on growth. We then entered into a period of time where everything's been focused on much more short-term profitability. We didn't let that change our strategy. We didn't say, okay, cancel all of our technology programs.

36:39we are focused on still building a very large company. And so I have to make, we have to like explain that to public shareholders that sometimes don't fully appreciate the breadth of what we're building or the scale of what we're attempting to build. But that's okay. Like if you sort of have, you go into building a car company and being a public car company knowing you're going to be a little misunderstood for a while. You're going to have deep convictions on things. Not everybody's going to share those convictions. I think that's a skill a lot of entrepreneurs pick up. It's certainly something I've learned.

37:10And like, I'm very comfortable with lots of people disagreeing with me. Like if you're someone that like is uncomfortable with people telling your ideas are wrong, being an entrepreneur is really hard. But if you're okay with people saying, I don't agree with you and saying, well, that's okay. You don't have to agree with me. That's sort of what you need when you're building large, complex businesses. Yeah, you're able to shrug it off. And it sounds like that's changing, by the way, with the profitability. Because obviously I did see your most recent earnings report. You guys had a great day with the stock there.

37:39I'm sure it was a good day. Of course, we have to be a profitable business. That's like, I say this with a smile, like the goal is to make money. We have to make money so they can self-fund our significant growth. But there's the reality of to get to that point and to get to that point where we're designing for scale, there's a certain level of investment that we've had to make in the business. Is there a timetable, just briefly, because I want to get to some of the other investments you've talked about. Is there a timetable you have in mind when it comes to that kind of steady profitability going forward?

38:13Because obviously you are still heavy in a lot of investments right now. I think the balance is always balancing growth with profitability. And today we're obviously focused on growth on a journey to get to profitability. given the scale of the business and with like the mass market nature of r2 and r3 that trade-off becomes much less challenging meaning we start to generate enough revenue that even with the growth mindset investing in tech and heavy investment in technology the revenue offsets all that op x and generates you know positive free cash flow and that's like getting to the escape velocity i think every especially every hardware business thinks about this and plans for this.

38:55And so we're like hyper-focused on it. And obviously the R2 is meant to unlock that further. I mean, that's a huge, that's part of the... The Rivian's business model doesn't work without R2. Like if there was no R2, we couldn't look the way we look. We couldn't have such a large engineering team. We couldn't have such heavy engineering budget. I want to get your take on AI, because in December Rivian announced it had developed its very own AI chip, something extremely rare for an automaker. I know you were very excited to unveil this. Why take this on, though, given the complexity required? It sounds like you had your hands full already, first of all.

39:28But, you know, obviously, I'm mindful of the fact that in recent years, a chip shortage exposed how fragile the supply chain can be, to your point earlier. Is this a hedge against that? Or in how much of this is about reducing your own reliance on companies like NVIDIA versus a bid to compete more broadly in the AI compute space? Well, in developing an in-house inference platform, it's not like something you wake up on a Sunday and say, let's go do this and it's done in a week. It's a big commitment. It's many, many hundreds of millions of dollars of time over years. You have to build a team. There's lots of iteration.

40:08You have to build a development environment around it. So it's a highly contemplated decision. The decision for us to bring our inference platform in-house, which we announced this, it's a very high-performance chip. So it's an 800-TOPS chip. We put two of them into a vehicle for 1 ,600 SPARSTOPS. So double that if you're talking about DENSTOPS, that joins of operations per second. The chip itself can process 5 billion pixels per second. So it's extremely high-performance, and it's optimized around vision-based robotics. And so that decision to do that was because we felt the most expensive part of a self-driving system, which is surprising because it doesn't get enough attention, is not actually the perception.

40:54We often talk about cameras and radar and LIDAR and the different perceptions. All those elements have become fairly cost-effective. It's actually in the brain. And so we wanted to have a very cost-effective way to integrate very high levels of self-driving because we have very high levels of compute built into the vehicle. And it's built around a deeper thesis that we have that the biggest shift to transportation, as big as electrification may seem, is not actually electrification. It's the shift to vehicles being able to drive themselves. We're at like the very beginning of this. So in terms of consumer vehicles, today there's what's called level two system.

41:35So it's where the vehicle, you know, like in a Tesla FSD, it can drive itself, but you still are paying attention to the road. A Waymo is a level four. It can drive itself empty. It doesn't need a person in a driver's seat. But our view is that over the next few years, we're going to see everything, like every type of vehicle, whether it's owned or robo-taxi, becomes something that should have level 4 capability, meaning it can drive itself empty. Now, whether that's in two years or three years or four years, you can debate, but certainly in the very near future. And so as we look at the remainder of this decade, we think consumer confidence in self-driving is going to grow a lot as they start to get exposure to early examples of things like Waymo.

42:17But by the time we get to 2030, 2031, there's two ways to look at this. Not having very high levels of capability in terms of self-driving will dramatically reduce market share. And the inverse of that is true. If you are among the small number of companies that have that capability, you can greatly expand market share. Or you could sell the technology. You could do one of those two things. But it's something that can be monetized through a few different paths. And so we made the decision to say, that's so important. If we believe, if we're convicted to being in the transportation space, we must have deep conviction here.

42:54It's a necessary element how this space evolves. And so we invested heavily in it. So we built, if you think of, to do self-driving, you have to have a lot of ingredients. So when we launched in 2021, we had a fairly off the shelf system. We used a mobile iFront camera. We launched in 2024, an in-house system. It's still using a video processor. but it was part of a broader effort to bring all of this in-house since we have perceptions in-house so our camera platform is in-house the compute with our in-house processor comes in-house with R2 and then the model which is trained very different than how they've historically been built but now it's built much more on like a neural net type architecture it's a trained end-to-end a lot of it's inspired from the transfer-based encoding that we've seen be deployed in LLMs So it's a very different style of building the model, the software side of this.

43:47And we're wildly bullish on what will happen with this over the next few years. Is the goal here to license or sell that out to other automakers or transportation providers long term and obviously continue to grow beyond the traditional OEM? I would say it's one of the major pathways to monetize, to have this make money. but I think the answer we have to recognize well there's a there's a deeply held belief we have which is in time uh if you're a vehicle manufacturer and you want to maintain market share we think it's a must-have that you have these two things the first is the vehicle needs a completely different software and electronic architecture than what has historically been in cars so it's going from a world where and this is true for every company with the exception of Tesla and Rivian, cars have many, many ECUs.

44:39So depending on the vehicle, we'll call it anywhere from 75 to 150, what's called electronic control units, little mini computers that run little islands of software that are supplied by suppliers. And all these ECUs are plugged together as part of a network architecture to essentially in aggregate deliver the software features in the vehicle. We could spend a lot of time on this, but for historical reasons, that grew out of the first ECU was associated with fuel injection. And over time, like a field of weeds, all these ECUs started popping up as features within the vehicle started to become software defined.

45:13As I said, like a field of weeds, it's almost the precise opposite of what you would design if you're architecting a system. And if you're architecting a system, you design what you see in a Rivian, which is a very small number of computers that do all the thinking in the vehicle. So more of a centralized computer, zonal architecture. And so the reason that's so important is you don't want to have your software distributed across a hundred different companies on a hundred different little islands of code. You want to have one software code base that you can do updates to and add features to. And so today what you see in all the sort of legacy vehicles is this system of many, many ECUs and they're very hard to make software updates.

45:55You can do it, but it's like brute force. It's the reason you don't see a lot of over-the-ear updates outside of a Rivian or Tesla. And so our view is that architecture is hard today. It's impossible in a world where you're imagining an AI-enabled vehicle. And so I think every car, if it wants to maintain market share, will have to shift to this software-defined architecture. So that's piece number one, but that's not necessarily binary. You can brute force your way, as I said, for a bit, but then eventually it just becomes so archaic that it's hard to hire people to work in these development environments.

46:30It's very hard. That's number one. Second thing is it's inconceivable to us that you'll be able to maintain market share without high levels of autonomy when we get into that, call it post late 2020s, early 2030s timeframe. And that will become also increasingly true. And so the reason I say all that is if we believe those two things to be true. Every car company in the world has three choices. The first choice is they could decide to ignore that and shrink so they could lose market share, which is a choice they may not explicitly make, but they may implicitly make by just their actions. But it's important to note that that is a choice.

47:11They could decide to develop those technologies in-house. That's very hard because for most car companies, these are not skill sets that are embedded within the businesses. The way many car companies grew up, they just grew up differently. They grew up as tech companies or software companies, so they don't have a lot of internal expertise around designing electronics, designing software, designing AI systems. Or third, they could go buy it from a third party. Our view is most companies are going to fall into the latter two categories. They'll either try to do it themselves or they'll realize they don't have those capabilities and it would take too long to develop those and they'll decide to buy it.

47:50And so we think there's pretty interesting business on both the software architecture, so the software-defined element of the vehicle, and then self-driving ultimately is going to be something we think most car companies are in need to buy. And so that's a path to monetize those technologies. The other path, of course, is just market share, selling more vehicles. And so we're going to pursue both. And so we did a deal with Volkswagen last year on the software-defined aspect, we did a$5.8 billion software licensing deal to provide the ECUs and the software architecture and software platform for Volkswagen Group products.

48:26So it's like Audi, VW, there's a bunch of brands. I think that's a very good example of an automaker that could use your expertise, by the way. And no shade to them, I'm just saying. I think that it's a perfect tie-up. Excuse me, second largest car company in the world, outstanding set of brands, but it's a really interesting partnership for us where we're able to deploy, really at scale, our technology across different form factors, different price points, different markets. Another venture I recently learned about is Mind Robotics, a new company that you started last year. First off, genuinely, how the heck did you find time to do this?

48:59And second, tell us how this kind of came together and your strategic vision here. We just talked a lot about AI. Vehicle autonomy is one form of AI in the physical world. we for a while have been convicted on ai manifesting in the physical world in many ways but industrial robotics is is another big enormous opportunity and i live this you know every day with rivian we have a plant that employs lots of people to build our vehicles and as we're thinking about scaling there there are real constraints on how many people are available to work in these plants and i think a real opportunity to take a lot of these tasks a lot these jobs that are highly repetitive and can be quite challenging and develop highly dexterous robotics that have reasoning capabilities.

49:47And so to make a very long story short, we've been looking at this for a while. And last year, as you said, I started another company and it took a little bit of time for myself along with Rivian's board to figure out how to structure that because I'm the CEO and chairman of Rivian and I intend to stay the CEO and chairman of Rivian. And I wanted to start and run this other company. Where it stands today is it's like enormously exciting. It's really at the, I think we're at a point in time where we're about to see robotics in many different parts of the world and many different applications. But we think one of the biggest applications is industrial robots that are specifically designed to do manufacturing and to do human-like skills in manufacturing.

50:38Well, so I previously covered, you know, business and tech out of Asia Pacific, and we've seen a lot more of that as well over there, I think, especially with some of the factory applications and whatnot. I'm curious if you're taking any learnings from, you know, other regions as you continue to build out this new venture. When we think about robotics and manufacturing, you can Google it and immediately see every car company in the world, every scale manufacturer, Rivian included, is using a lot of robotics as it is today. But these are very simple, PLC-controlled, six-axis robots that are doing fixed motion planning with very specific end effectors.

51:15So they're not generalized robots. They're not doing human-like skills. They can do things very fast and very repetitively. But when you introduce variation or when you introduce human decisioning, it's where that type of robot really doesn't work well. And so the concept of doing robots that have human-like capabilities or even human-like form factors is something that's been around for a while. But it hasn't been really until recently that the ability to create a brain for these, a model for these to operate with, has really facilitated the realistic deployment of these at scale. And so it's something we've been watching very closely and the way we're now creating these VLA models and these world models to describe how to behave in a plant with the benefit of lots of data feeding them is enormously exciting, but it's very new.

52:09This is like not something that was around more than a couple of years ago. And I think precisely because of that, it's what gives me so much confidence that this is like very realistic to deploy at scale. Rivian is a great first customer. By virtue of Rivian being a shareholder in mind as a customer, it allows mind to focus on a scale deployment as opposed to demos. And a lot of the robotics world for a while has been in this loop of lots of demos and lots of proof of concepts, but not like how do we deploy 100 robots? How do we deploy 1 ,000 robots? Lots of trade shows and whatnot. So we're very focused on execution.

52:48Yeah, that you're building robots that will perform real tasks, real plans at real scale. What's like one example of that? What's something you really just want them to be able to get up and do? So like hard tasks or things like routing a wiring harness and plugging it in. easy tasks of which there's millions and millions of these types of tasks across manufacturing globally is something called pick and place it taking a a part let's say in a box that's unsorted like a box of parts and placing it into like say a shortage a sorted shelf or into a fixture and so it doesn't require enormous dexterity or manipulation it's a relatively straightforward task, but it's one for which there's enough variation in the process that classical automation doesn't work for it.

53:35It's where we see lots of humans doing this type of work, highly repetitive. In some cases, it can be quite tiresome or cumbersome. And so this is one of the very immediate things we can address with MIND. But when I think of the capability of MIND, it's not going to end at pick and place. It's going to include every type of manufacturing. So dealing with objects that can be manipulated. So these are like soft goods, carpet, headliners, wiring harnesses, dealing with connectors that have force feedback when you plug them together. These are a lot of the skills we're both training against and then ultimately building the dexterity and the mechatronics to deliver on those tasks.

54:17And zooming out, RJ, what is your endgame here? Because you have your hands in so many different things. I'm just curious if we were to sit back down again and let's say 15, 20 years, where would you really want to be? You think of like the set of companies. So Rivian is obviously like working very hard to help advance smart, connected, increasingly autonomous vehicles. And, you know, very soon our view is that vehicles, a modern vehicle should just be capable of driving itself. So it should be capable of going to the grocery store, picking up your stuff, dropping you at the airport. and we think that'll be of course in robo taxis but importantly in just the car you own your car you own can be running tasks for you and we're developing the technology to be operating at scale producing many millions of units a year for that to help on that journey mind was created to build the robotics to allow us to get the cost structure of these you know reduced dramatically and so if you look at the labor content that exists in a vehicle not just in the vehicle assembly plant, but across its supply chain, it's quite high.

55:22And so allowing us to build really compelling products at a much, much lower price point is something that Mind's working towards. And Mind, of course, is designed to support all types of manufacturing. So building cars, microwaves, soft goods, clothing, everything we see is the surface area that we were working to address with Mind in manufacturing. And then there's a micromobility company that was a spin-out from Rivian. And my goal there, the name of the company is also. And the name is sort of captures what its purpose is. And really it's bolted on this idea that Rivian is working very hard to electrify and create incredible products in the vehicle space, but we also need to rethink and electrify everything else.

56:13And so it's the name also is working on everything else. So that's all these different form factors that are smaller than a car in many parts of the world are primary modes of transportation. In the U.S., we're so used to it, but our primary mode of accessing personal transportation is a car. So there's over 700 cars per 1 ,000 people. That's two or three orders of magnitude more car ownership per person than what you have in most countries. And most countries are using two-wheelers or three-wheelers to get around. In some cases, like quadricycles or four-wheelers also is working to rethink that segment in a really meaningful way with both technology connectivity obviously electrification and autonomy in these much smaller form factors so stay tuned that's an exciting one i'm eager to see what you do with that thank you so much rj now we're going to wrap with a quick rapid fire all right okay don't overthink it just say whatever comes to mind ready go to playlist in the car go to playlist uh i'll be like electronic music oh Interesting.

57:12Peaceful. I would have like picture like a country music star for you. Okay. Tesla or Toyota, who keeps you up at night as a competitor? I wouldn't say. You know, the classic line of focus on what you're executing, not looking to your left or your right. Favorite way to unwind after a long day? Usually reading, reading something, reading articles or. What do you like to ask when you're hiring? Usually tie it to whatever the role is to a specific example, challenge or problem to see how one would construct putting that together or solving it. One piece of advice you come back to often. I'd say zoom out.

57:49Take a step back and look at the full system. I love the simplicity of that. RJ, thank you. I've enjoyed this. Thank you. Thanks for joining us on Behind the Business. For more stories on business and leadership, follow us on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts. See you next week.