Great. A lot to cover. We'll get right to it. Maybe just to kick off, RJ, since we are at a consumer conference, talk to us about what's led to Rivian's success in such a difficult industry where so many other kind of EV startups struggle to build a brand, to build demand. Kind of what set you apart?
Yeah, I mean, uniquely in the transportation automotive space, you've got a product that's really complex in terms of the number of decisions and the number of components that must come together ultimately to make the product. Just doing that isn't sufficient. The combination of attributes, positioning, design, of course, price point needs to really connect with the market. Not only the product-market fit, but the result, the aggregate result of all those features needs to come together on a brand position that's really compelling. We spent a lot of time as a company early on debating what the positioning of the business was going to be.
We had a couple of pretty big pivots and ultimately landed on this really extreme point of clarity around creating a brand that helps both inspire but also enable people to go do the kinds of things you want to take photographs of. That brand position has proven to be really elastic in not only supporting our flagship launch product line, but supports what we call R2 and R3, which are follow-up, much lower priced products which we're about to launch next year.
Awesome. Tell us about Rivian's customers. Are they first-time EV buyers? Do they tend to come from the same segment? Do you pick up in SUVs? As you launch R2 and then R3, how are you thinking about that customer profile and how it might change?
Yeah, when we started to think about the product and the position of the product, we did a bunch of studies that would look at what was our ideal in terms of customer demographic. What we'd hoped would happen was we'd have a broad spread of customers that come from different segments, different political backgrounds, different cultural backgrounds. Of course, we had folks we'd hoped to have folks who would come out of non-EVs, which would be to say that Rivian creates new EV customers. That has proven to be true. The vast majority of our customers have not owned an EV before, which is great. The other thing is we wanted that first-time EV experience and the first-time experience with Rivian to be really remarkable.
Coming back to your first question on brand, there's a lot of ways to look at the strength of the brand, but one of the helpful ways is to look at how externally validated folks look at this. Consumer Reports does an annual brand survey where they actually survey customers of every brand and every company. For the last 2 years in a row, we've come out by far as the number one rated brand across all auto companies. This past year, I think we won by something wild, like 13, 14 points. It was a significant margin between us and next closest, and we had a similar level of success in the year prior. That, of course, translates to market share. Our flagship products, which we have today, are the dominant market share players.
If you're looking at an electric SUV over $70,000, we have a commanding level of market share around 35%. Our ASP is much higher than that. Our ASP is around $90,000. The hope is we can translate the brand success, the customer satisfaction, and the market share penetration we've had at this relatively narrow TAM, very high price point, into the much more affordable price point with our R2, which starts at $45,000.
Terrific. I think last year at the Investor Day, you shared that brand awareness in the U.S. is only about 40%. Any updates on that metric? Kind of where are you seeing that trend?
It's continuing to grow. I mean, for all the obvious things, we have efforts to drive that through marketing. We have some endorsements we have or partnerships we have with specific brand advocates. One of the most powerful things is just having more of them on the road. If you're in Southern California, Northern California, Seattle, Portland, increasingly here along the East Coast, you're seeing more. On the West Coast, they're everywhere. A good metric is my drive to my kids' school is about 10 miles. 2 years ago, we'd count about 4 or 5 Rivians on the way in. Today, every trip is over 25 Rivians. Of course, we're in the Palo Alto area, which is a very dense Rivian population.
We're seeing them more on the roads, and it just helps drive overall awareness.
Terrific. Maybe, RJ, on that point, because I think you raised an important point, when we look at the EV market today, it's kind of been very uneven from a geographic basis. When you're thinking about the R2 and R3 launch, could those potentially catalyze demand in the kind of central part of the country? Are there things you can do to help promote that either in product or in brand?
Yeah, I mean, just to call it out, I mean, like Santa Clara County, EV purchase rate is like over 40%. Across the whole United States, it's around 8%. You do see this real clustering of EV sales, particularly on the coast, as you called out, the center of the country a little less so. Part of this is just due to lack of choice. In the combustion world, there's on the order of 300 different nameplates you can pick from. That's brand and product line. A lot of those are very compelling. In the EV space, under $50,000, which is where the majority of the market is, there's very, very few. I would argue, honestly, well under what fingers on your hand. It's like a couple. Of course, you see that manifest in significant market share accumulation with one company, with Tesla.
Tesla's Model 3 and Model Y are exceptional products, but it is one specific point of view as to what a vehicle is in terms of form factor, design aesthetic, brand positioning. To go from 8% to 20% or 30% or ultimately 100%, we are going to need more choice than just one compelling option. We need to see market share much more distributed and much more players in the space.
Yeah, actually, I couldn't agree with you more on that level of choice. Maybe one related question to that is, as you're planning for R2 and then R3, how are you thinking about the number of trim combinations you would offer? Because on one hand, if you go with few trims, it's maybe more efficient for your manufacturing. Maybe it's better for margins in theory. Then maybe you do limit the choice of having some design changes. How are you thinking about that level of choice within your next kind of platform to try to widen out the buyer base?
Yeah. On R1, we had a lot when we launched, we had a lot of trim combinations, too many in terms of color combinations with powertrain and battery pack. On R2, we've really narrowed that, but we've done it very thoughtfully by combining and putting essentially into baskets different sets of features and different combinations of color and interior trim. What that's allowed us is to dramatically reduce the SKU count and simplify logistics beyond just the production side, but actually simplify the movement and placement of vehicles for sale. One of the unique things about being a direct-to-consumer brand, Tesla, of course, has this as well, is we have to be very thoughtful around where we place inventory. We don't have a traditional dealer system that absorbs inventory and therefore absorbs working capital. It's on us to own that.
We try to minimize that as much as possible. Customers want to, you walk in, you see a vehicle, you like it, you like to take delivery of it that day or within the next few days. Simplification of that portfolio actually helps with the delivery timeline. We are talking about low hundreds in terms of number of possible build combinations, but that is an order of magnitude less than what we have on R1.
Terrific. As we start production kind of next year of R2, how soon thereafter should we expect a $45,000 version? It's always tempting to kind of go high initially, low volume, high ASPs. The pushback to that is that maybe then consumers start to think of the vehicle as priced higher than it actually is, more in terms of volume.
I mean, we talk about this all the time. We're going to have a launch configuration, which it's like balancing. Everybody wants something slightly different, but we want to have a very narrow set of build combinations that we start with to simplify logistics, simplify production. That initial narrow combination of vehicles we're building will be higher priced because it's a fully loaded vehicle. Shortly after launch, we'll introduce different specs, but importantly, our $45,000 spec. It'll be pretty quick. Now, the big challenge we have on R2, which is similar to R1, is there's a really large demand backlog. R1, we really didn't get this right. You have a bunch of customers. If we say we're starting production on X date, everyone, regardless of when you put your order in, believes that you're getting your vehicle on that day.
It is not as if we can instantly deliver hundreds of thousands of vehicles on day one of production. There is going to be a long period of time for some folks to get their vehicles. It is just expectation management. We are going to be doing a lot over the next year to manage folks' expectations to say, "Look, even if you have a reservation, you may not get your vehicle until 2026," which is a great problem to have, but it is truly a challenge when you have this much of a built-up demand backlog.
Terrific. Maybe shifting back to the near term, any updates on just the current demand environment? It's been a bit of a choppy start for the EV market. How are kind of things trending?
Yeah. For us today with our R1 products, with the R1 team, the R1 asset truck and an SUV, those being premium flagship products, there is the inherent challenge of that segment being more compressed, where you see consumers maybe deciding instead of buying a $90,000 car to buy something that's more affordable. While the overall volume is still strong for the industry, the volume at the segments we're operating in is becoming more just more compressed. Now, our share within that segment, as I said before, is very strong. We have the highest share in the segments that we're in. We really look at this and say, "Boy, we wish we could have R2 ready," because the price positioning of R2 between $45,000-$55,000 really hits the bull's eye of where the market sits today. The average new car price is around $49,000.
It's a really nice positioning. The product packaging and overall layout is just perfect for the market.
Great. I think one of the interesting things related to that that came out of your Q1 earnings is I think you had record demo drives in Q1 despite the just overall more challenging kind of sales environment for the industry. Do you think that record demo drives are a potential indicator of pent-up demand? When the macro picture improves, maybe you do kind of see a bit of a snapback just because people are still engaged very much.
are so many variables right now that influence consumer behavior. We have seen a level of volatility just in daily order rate that has been very unique to the last, call it, 6 months, just because of the overall macro concerns around what is going to happen to the economy. You see folks as if trade relationships, if tariff discussions go a certain direction, we see demand move in different directions. If we see the perception of how some of the tax credits will be interpreted or enforced or changed, we see big swings in behavior. In many ways, I think the highly volatile nature of, we have had situations where we have had the best day ever in terms of orders followed by one of the worst days ever, which makes no sense. It is just the swing of the news is driving behaviors of consumers.
I think that's probably going to continue until we see things stabilize with just the overall trade environment.
Terrific. Maybe one more near-term question, then we'll get into the autonomy platform, which is very exciting. Any updated thoughts on tariffs? There's been some concern in the industry around rare earth. Any updated thoughts around those topics?
Just as context, the rare earth metals issue is a big issue, not just for automotive, but many industries. Specifically in automotive, to make the electric motors that are used in every vehicle on the road today, whether it's us or Tesla or someone else, they're using magnets. Those magnets have heavier earth metals that go into them. If you were to do a Google search and say, "Where does the world's dysprosium, which is a heavier earth metal, where does that sit?" you say, "There's lots of dysprosium. It's in many countries and many places." The challenge is that by far and away, the vast majority of the processing of most of these materials resides in China.
As the trade tensions between the United States and China rose, China put in place export controls that effectively limited the export of these rare earth metals to the United States. That was a huge concern, surprisingly not talked about very much because it was not just electric vehicles, but all medical imaging equipment, lasers, a lot of very important industries, the defense industry, frankly, would just be shut down without rare earth metals. We spent a lot of time figuring out how we would solve that in, call it, less than elegant supply chains to be able to continue production. Fortunately, as the trade tensions with China have eased, those export controls have also been eased. Those rare earth metals are starting to move again, which is really important.
It is a reminder, I think, from an industry point of view, from a supply chain point of view, of some of these really critical dependencies that we have on other countries, and here specifically China. We see a heightened focus right now on mid- to longer-term solutions to either remove heavier earth metals from the design, which are new technical solutions, or to find new ways to create those processing or downstream supply of those materials, which is not something you do in a matter of months. It is something that will take many, many years to do. There is, I'd say, renewed investment interest in that space.
Terrific. Shifting to the autonomy platform, obviously, a big part of the Rivian story as well. Maybe provide kind of your view of what the autonomy platform can be in the next few years, how you're approaching this megatrend maybe differently than some of the other companies out there.
Yeah. I mean, maybe a little bit of history on autonomy because it does all link together. If we think, wind the clock back to, call it, 2014, 2015, it's when we start to think about vehicle autonomy or the idea of a vehicle driving itself in a more substantial way. The way those systems were developed is you would have a perception stack, so some combination of cameras, maybe radars, maybe lidars, but a set of sensors that is perceiving the world that would take in imagery or take in information and process it. They'd identify objects. We would then classify those objects, and those classified objects would have vectors associated with them, so velocity and acceleration in X, Y, and Z. All that object-based information would then be handed to a rules-based planner.
When I say rules-based planner, a planner that has a set of rules that makes determinations around what to do based upon a pre-programmed set of intended behaviors for all those objects. That approach is what you saw Waymo launch with. Of course, Tesla Autopilot was built with that. What you'll now hear that called is you'll hear people say that's AV1.0. It is a rules-based program solution. Just a couple of years ago, in the late 2021, early 2022 timeframe, as we start to see the use of transformers and full neural net approach to language emerge with the idea of these large foundation models and large language models, that same approach started to be adopted for autonomous vehicles.
What has dramatically shifted how we develop these systems, including ourselves, our Gen 1 vehicle that we launched in 2021, was a rules-based solution. It's not the right technology topology for the long term. We launched our Gen 2 about this time last year, which is completely designed around AI. To be completely designed around AI, to define or to describe what that system looks like, different than having a bunch of perception that then classifies objects and associates vectors with those and hands them to a rules-based planner, you now take all the data from your perception stack. Before making any decisions or any determinations on what those things are, you fuse that information together to create a very accurate and robust view of the world. You then take actions on that.
The actions are informed by a large-scale, large-parameter foundation model that is built through a data flywheel that you are training by observing the operations of the vehicles in these different conditions. You create a true neural net, a true model of how to behave in the world. Much like the way we as humans would learn to drive a vehicle, we have now taught our AI systems to do the same. The reason I call that out is there is truly zero carryover between an AV1.0 and an AV2.0 solution. The ingredients necessary to be successful in the long term are very different than what was important to be successful in AV1.0. What you need to be successful now is you need complete control of the perception stack, meaning your cameras have to provide raw information.
They can't go through any sort of filter or any sort of layer of abstraction, let's say, through a tier one supplier. We need raw information from your cameras. If you have them in your radars and if you have them in your lidars, you need to build a robust view of the state of the world real-time in the vehicle with your inference platform. That data as well needs to be piped offline through a really robust data platform that forms this really large data flywheel that you're using to train offline this model. This offline model is a large parameter. It's a multi-billion parameter model that is understanding and building an understanding of how the physical world works. What's running in the vehicle on inference is a distilled version of that. It's a smaller parameter model, but it can run real-time in the vehicle.
To do that, you need control of your sensor stack. You need control of your data platform in the vehicle. You need control of the inference platform in the vehicle. You need a ton of GPUs. You need a really efficient way to move data between the vehicle and the cloud offline. You do not want to do that through LTE because you will spend a fortune. You need to have a set of vehicles that have really robust Wi-Fi policies and users that have selected to enable Wi-Fi and a whole bunch of incentives to drive that to ensure you have a low-cost way to move all this data. There are very, very few companies that have architectures like that. Of course, Tesla is in that camp. We are also in that camp.
For this system to work, you'd benefit from the scale of the flywheel, meaning you need a large enough car park for it to start to create some of the momentum. On our Gen 2 vehicle, we architected everything, contemplating this. We launched it on R1 middle of last year. That fleet size has grown. The data flywheel has become very capable. We're now starting to see the beginnings of this really non-linear growth curve in terms of features and capability. We've announced we're having an autonomy and AI day in the fall. We want to really pull the sheets off of everything that's coming in terms of pipeline. I think this is the most important area of the business. It's our biggest by far spend category in R&D.
I think there's a lot of misconceptions around how to design these systems, what a true data flywheel looks like. When I say that, how you architect the dedicated cluster of GPUs, what your inference platform needs to do, and how it needs to be architected, there's a lot of very different strategies than what we saw for 10 years prior. I couldn't be more excited about this part of our business. We're just now starting to see it's like classic. You spend a mountain of time, effort, and money on all the stuff below the surface. Finally, it sort of comes up above the surface. You start to see it in the form of customer benefit. We're just starting to see that now.
By the fall, customers will start to get fairly trained on this rapid rate of progress on the features in the vehicle. That'll coincide with when we have our AI day.
Terrific. Archie, when we think about Eyes Off, where everyone's trying to get to, what do you think is the killer app for consumers? Is it highway? Is it sharing robotaxi? Is it other things that the vehicle could do? There's so much.
Yeah. This is really an important point on this new generation, this AV2.0, as it's often referred to, which is in the previous world, you had a 2-pronged approach for how you, in AV1.0, you had hardware light systems that sort of capped out because of a perception stack at a level 2 feature, so hands on wheel, eyes on road. The other approach or topology was to have really hardware heavy, à la Waymo, but tens of thousands of dollars of perception and compute in the vehicle, but that would raise the ceiling, if you will, up to a level 4, hands off, eyes off, occupant or driverless vehicle. What's happened as you move to an AI-centric approach is it's the same stack. It's the same architecture.
It's just a continuum of the robustness of the model, the robustness of the perception stack to identify and see challenging corner cases, and the robustness of that sensor stack in informing and growing and strengthening the model. For us, we look at what we're now developing as being on this continuum where there's not an artificial or glass ceiling that exists. It truly can scale to level 4, but it will take further increases in compute on board, so inference, further improvements in perception, and a lot more training of the model. The model is accretive.
The training that you do with a lesser sensor set informs the brain, if you will, the neural net in the same way that we as humans, as we develop better senses and grow as humans, our brain does not throw away the information we gained from ages 1 to 5 to replace it with better information we get from ages 5 to 10. It is accretive. It is aggregating together. I say all that because I think there are a few big breakpoints from a revenue point of view or go-to-market point of view that are really important. In a personally owned vehicle, we think one of the most valuable is getting to highway hands off, which we have done today, but then hands off plus eyes off, which means without needing to, you do not have any need to look at the road. You can be doing emails.
You can be reading a book. You can be having a conversation with your kids in the back seat. You are not an active participant in any way whatsoever in the operation of the vehicle. That hands off, eyes off, you might hear called level 3. You are in the vehicle. You might be asked to retake control, but you are given 20 seconds to retake control. It is not a dynamic active dynamic retaking of control. I think that naturally makes a lot of sense on the highway because you have large chunks of time. Over time, that becomes valuable everywhere. Our view is we go hands off today. We go hands off. We add eyes off fairly soon.
We then go hands off, eyes off beyond highways, so inclusive of your feeder roads, your surface roads, surveillance environments, which is sort of full scope level 3. The natural step beyond that is to remove the need for a driver in the vehicle so the vehicle can reposition itself or move itself when it's empty. All that exists along a common technology backbone where everything that's being built is building towards that end state.
Interesting. Archie, as you build the autonomy platform and launch more features, clearly the technology can also accrue to vehicle safety. Can you talk about in a couple of years, how safe are Rivian's vehicles? Are there opportunities for you with an insurance, reduced cost of ownership? How big could that be for the company?
Yeah. I mean, to be clear, if you use Rivian's what we call highway feature today, where you can get on the road, it drives itself on the highway. It is remarkably safer than a human, than the best human. We have an insurance offering. One of the reasons we built the insurance side of our business was contemplating exactly this. Your rate actually goes down the more you use our Driver+ feature. If, let's say, you use it all the time when you're on the highway, you will have a notably lower rate because the risk profile is dramatically reduced. I think this will just be a common trend. It is all the obvious things. We as humans, we're good at dealing with complexity. We're actually quite bad at dealing with boring driving. We end up texting.
We end up looking away. We end up getting distracted. If you look at the number of deaths in vehicles today, it has not really improved that much despite the amazing progress we have made in passive safety, the ability for a vehicle to absorb a crash because the number of distracted accidents has gone up so much. I do think we are at, sort of call it, peak automotive death rate. I mean, that is not a very glamorous way to describe. I think we are going to start to see that start to come down dramatically as we see humans play less and less of a critical role in the operation of the vehicle. With that, you will see insurance rates come down. The challenge with autonomy is it is not like you being 10 times better is enough.
Just to put some scale to this, in the United States alone, which is a relatively safe driving environment as far as countries go, we have over 40,000 deaths per year in vehicles. If every car instantly was driven by the car itself and not the driver, and we said there was an order of magnitude of 10x improvement in safety, we would still have 4,000 deaths, which is nowhere near good enough. You can imagine how hard it would be for us as businesses to operate if we had that many people dying because of mistakes of the vehicle. It needs to be orders of magnitude better. We think of it as 4 orders of magnitude better. That will come.
I think we'll see that very rapidly in highways where the idea of a vehicle crashing on a highway when it's in an autonomous mode will be near zero. Then we'll translate that into these more complex urban and suburban environments.
Terrific. You mentioned that the hands-free eyes-on feature you have today, safer than a human. Was that performance better or in line with what you were expecting internally when you launched it? I'm just curious if the inference and the training is actually doing better than what you.
The metric for us, the acceptability criteria for us to say this is a feature that we're going to offer, it had to be meaningfully better. I would call it a requirement. We wouldn't have launched it if it wasn't better.
How's customer reception been?
Great. I mean, this just blows my mind. Today, on our vehicles with a narrow set of applications, meaning highways, it's not every road, a very significant portion of our miles, double-digit % of our miles overall are driven by Rivian, not by the driver. You can imagine as we grow to every highway, every surface road, every suburban street, we think that once you use a platform like this, it's really nice. When you get your time back, I think most often people will choose to say, I may enjoy driving, but most of the time, I'd rather just have the car drive me.
Interesting.
We think that it's a very sticky take rate.
Wow. I know we're out of time, but I'll sneak one more in, RJ. Maybe early days in the autonomy platform, but any thoughts on licensing opportunities in the future at least?
Yeah. I mean, we have quite a bit of experience in working with other companies. We have a very large partnership with Amazon. Amazon's also our largest shareholder. We build, if you see their electric vans, I do not know if they're operating here in the city, but they've got the largest deployed fleet of electric delivery vans in the world. We actually design and build those in partnership with them. More recently, we have a large $5.8 billion software licensing deal with Volkswagen where we provide them with our base operating system and some of our core compute platforms. I think that certainly as the auto industry moves to become much more dependent around software and technology, and in particular AI, traditional car companies do not have these competencies. It is not something that they've built.
We see a lot of potential business in providing some of this technology to other manufacturers.
Terrific. I think we'll leave it there. It was a great conversation. Archie, thanks so much for joining me. Thank you.