We're very excited to kick off this next discussion with Rivian and with RJ. Rivian is an organization that has achieved something that very few have been able to do in the recent history of the auto industry, which is to start a car company from scratch and build a brand that resonates emotionally with consumers. In that process, the company has really chartered its own course, vertically integrating in key areas including software, electronic architecture, distribution, and EV charging. In doing so, Rivian has demonstrated clear competitive advantages, especially in software and architecture. It has also clearly been demonstrated again with the recent JV with Volkswagen, where the two companies will co-develop a next-generation platform leveraging much of the work Rivian has already done, including on the recently redesigned R1. Now looking ahead, execution over the next few years is going to be very important.
Rivian's mass-market midsize SUV called the R2 is launching in the first half of 2026. Construction on your second plant in Georgia is also set to begin in 2026, and management expects to reach positive EBITDA by the end of 2027. Very much a lot on your plate. Here to discuss these targets and the evolving EV landscape is RJ Scaringe, founder and CEO of Rivian. RJ, thanks so much for being with us.
Yeah, thanks for having me. We're excited to be here. As you said, there's a lot of things happening within the business, so I'm looking forward to discussing them with you.
Yeah, same here. Maybe just to kick things off on some of the shorter-term dynamics and demand trends in particular, nearer term, you talked about Q1 deliveries of just 8,000 units as the quarter was being impacted by seasonality and unique demand challenges, including LA following the fires. At the same time, you're planning to produce 14,000 units, so it implies a pretty sizable inventory build. You are expecting 2025 deliveries of 46,000 to 51,000, which is down 1%-10% year over year and partially impacted by planned downtime. What, in your view, is the true underlying demand this year, and what tools are you using to keep it healthy?
Yeah, I mean, we're, as you said in your opening remarks, we're really pleased with the strength of the brand that we built and the excitement that we have for the products and not just R1, but also what's coming with R2. As we look at the remainder of this year, we intend to continue building on that brand strength. What we've guided to reflects sort of the view that with R1, given its high average selling price, the, call it $46,000 to $51,000 guide represents what we think is appropriate for demand this year and very realistic and, of course, achievable this year. In terms of Q1, the level of production being higher than what we're selling in this quarter is really us anticipating the shutdown that's coming in the second half of this year and wanting to build a healthy inventory level.
The other thing that's happening is we're seeing a pretty profound shift from how we sell vehicles, going from a world in which our vehicles are produced, they're linked to a customer at the plant, and then the vehicles are then sent to where the customer is, to a model that's much more inventory-based and needing to have inventory close to where customers are. As we think about customers coming in to buy a vehicle, they want to make the decision to buy the vehicle and then have the vehicle within 24-40 hours. We are also building out inventory within our distribution network to support growing sales. We've seen demonstrations of just how effective that is in Q4, and we're seeing that again as we speak now.
Now, there's been a lot of discussion around the Tesla brand recently and certainly some indications that maybe it is struggling a little bit. That could presumably create an opening for Rivian. I'm curious, have you seen any signs recently of either an uptick in orders or even indication of interest from current or prior Tesla owners in the last three to six months?
Yeah, I mean, when we look at the overall landscape for what customers have in front of them for choices, particularly when you look at price points under $50,000, there's very few highly compelling choices. I've said this a lot in the past, but if you're buying an electric vehicle and let's say you want to spend $40,000 to $50,000, there's really maybe one or two highly compelling choices, one of those, of course, being Tesla with the Model 3 and the Model Y. The lack of choice essentially creates a real glass ceiling on how much EV penetration we're going to see. Notwithstanding your question and just some of the sentiment that we're seeing emerge around Tesla, there's a massive amount of untapped demand that exists under $50,000.
I look at R1, the R1S, it's the market share leading electric SUV in the premium segment. Vehicles priced over $70,000. Of course, our ASP is quite a bit higher than $70,000. You can back it out from our numbers, but it's on the order of $90,000. That's a significant amount of money. The size of that market is relatively limited. Even if we're number one in terms of market share and we've got incredible brand appeal, Consumer Reports has an annual brand survey. We've come out for the last two years as the number one ranked brand and the number one rate of repurchase. There's just not a product for where most customers are, which is under $50,000.
I couldn't be more excited about what's to come with R2 and the benefits that R2 brings to the whole business in terms of increased scale that drives costs down not just for R2, but it also helps the cost structure on R1. We're ecstatic about what's to come with R2.
Maybe we can just touch on sort of the topic du jour, which I guess is tariffs. Maybe even we can expand it a little more and think about IRA and emission regulations. There's a lot of moving pieces, and we'd love to get your take on a few of those. Just big picture, how are you and the team navigating through this environment where it seems like policy decisions are changing quite rapidly and each of which could have a pretty meaningful impact on the organization?
Yeah, I mean, there's a few categories to consider here. In terms of policy, there's, of course, what we're seeing happen with the IRA. Our view is we are going to see that diminish over time. We can debate when and how and to what degree, but I think it's a fair assumption to say that that's going to reduce over time. That's been built into our plans for a while. We've recognized that even in the guidance we just talked about in terms of 2020, in terms of 2025 numbers. The more challenging long-term shift, macro shift that we're navigating now is the potential of increased costs for parts and systems coming out from outside the United States. Very long-time trade partners between Mexico and Canada have had billions and billions of dollars of investment in production capacity in those countries.
This is not unique to Rivian. This is across every manufacturer in the United States. If you're building cars in the U.S., you're leveraging to some degree and generally a pretty significant degree a supply base that's built across NAFTA. With that said, we're, of course, responding to it. It's very dynamic. We're with a number of different contingency plans that allow us to adjust. Given that most of R2 has been sourced, while these were discussion areas, a lot of our sourcing contracts, so the vast majority of R2 has been sourced. It's close to 100% of the vehicles sourced now, but that was done over the last roughly half year. In that timeframe, we built contracts that had some protections for us in the event of tariffs going up. We're, of course, now utilizing some of those protections.
I do think if tariffs for goods outside the United States are going to go up, we're going to see the price of vehicles either go up or certain manufacturers, more so than Rivian, have a hit on their margin structure.
Maybe just thinking about the longer-term implications, tariffs to some extent, I think you kind of alluded to that. If we see the IRA get repealed, if regulatory standards for emissions are rolled back, can Rivian still get to the long-term targets like positive EBITDA by year-end 2027 in your view?
First, we should separate immediate, medium, and long term. In the immediate term, if a 20% tariff goes in overnight, there's nothing we can do the next day to shift our supply chain. We may have contractual protections that reduce some of that cost to us, but there's not a lot you can do. In the medium term, we can work with our suppliers to find production locations that are more advantaged from a tariff point of view. In the long term, this will all settle out. We'll optimize our supply chain around the tariff structure. The big challenge we have today is there's just so many unknowns, and it's changing so dynamically and so often that it's very hard to make large long-term commitments from a production location point of view, knowing where it is today.
Now, with regards to R2, I do want to just say, because a lot of these discussions around a more protective tariff structure are already happening, when we source R2, it's a very USMCA-centric supply chain. It's already contemplated a lot of what's happening today. We made those decisions going back to summer 2024. We think we're relatively advantaged as we compare to other manufacturers in this way. If the consumer-facing credits go away, I think that in the short term maybe creates a little bit of turbulence, but in the long term, the shift to electrification is going to happen. The key for us is it's strangely become a topic of debate, like is the world going to electrify? Of course, it is. The entirety of every vehicle produced on the planet will be electric.
We can debate whether that's in 10 years or in 15 years or in 20 years, but it's going to happen. Of course, I tend to believe it's going to happen quicker than some of the more bearish points of view on this. Now, as that happens, we need to have choice. When you look at the market as it is today with Tesla commanding around 50% market share, that's reflective of a market without choice. It's reflective of a market with a singular great option. We badly need choice. We need choice that looks different than each other, meaning R2 is intentionally very different than a Model Y. It's a similar price point, similar size, but they couldn't be more different in terms of how they present in terms of design, in terms of features, in terms of content, the overall brand presentation.
That's a very good thing. It's not to say one's better than the other. It's just to say that it gives customers a different perspective. The scale that R2 brings will help us not only improve profitability or help mitigate profitability on a much lower bill of material cost structure and a much lower non-bill of material cost structure, but that translates immediately over to R1. There are a number of suppliers that are supplying both R1 and R2, and the capacity and volume we add for R2 helps provide another step change in cost structure on R1. The non-bill of materials COGS, that's the conversion costs, inbound logistics, everything associated with building the vehicle that's not bill of materials, not parts. We get a huge advantage of being able to spread those costs around more volume.
We'll see another benefit to R1 that comes with R2 there. Claire and I have guided to say 2027 is going to be positive EBITDA. We continue to maintain that guidance. We're very bullish on that. These are relatively deterministic things, meaning we're not wishing for this or hoping for it. We've negotiated our bill of materials. It's nearly 100% negotiated for R2. We can contemplate medium to worst-case scenarios for tariffs. We very clearly understand our fixed cost structure for running Normal, from running our plants to Normal, and understand what additional volume will do to spread those fixed costs across more units. We maintain a lot of confidence there.
Maybe just before we dive deeper on the R2 economics, I wanted to briefly touch on the R1. You launched the second-generation version in the second half of 2024, Q4, I think the first full quarter with that version. You had previously noted there was a real focus on costs, including material costs. There were hundreds of engineering design modifications in that new gen. At a high level, obviously, the cost of goods sold per unit fell considerably. Part of that was DNA, and the contribution margins remain negative. How should we think about the road ahead for R1, and where do you see the biggest opportunity to further improve the return profile?
It's an important question and a great question. If we look at Q4 of 2024 relative to Q4 of 2023, we took about $30,000 in COGS out of the vehicle, just over. As you pointed out, a big component of that was we resourced more than half of the bill of materials. We brought in new suppliers. Along with those new suppliers, we also made changes to the component designs. The one we've talked about a lot is the shift to a consolidated set of ECUs where we went from 17 in-house ECUs in our R1 Gen 1 to 7 in-house ECUs in R1 Gen 2 and took out a tremendous amount of wiring harness. That extends into we integrated a heat pump. We completely redid the battery pack. We updated the drive units, brought everything completely in-house on the drive units.
I mean, it's just like changes across the whole vehicle. That allowed us to take the cost out that I just referred to, that $30,000 in COGS. Now, there's further changes that are coming. There's further improvements that are going to be happening. Quarter over quarter, we'll continue to see those on the bill of materials. I talked about it a moment ago, but a further step change is the volume increase that we see with R2. The economies of scale that R2 helps provide to the business isn't just important for R2. That's a foundational element for R1. I'll take even like the center information display screen. We're using the same supplier between R1 and R2.
You can imagine there's a lot of cost savings that we can negotiate on the R1 product because we're also sourcing R2 for a similar form factor and similar screen. That's on the BOM. The other big driver here is the non-BOM items. We're sharing our paint shop. We're sharing our stamping operation. We're sharing a bunch of the management infrastructure and quality infrastructure in the plant. The ability to share not only that CapEx, which ties to depreciation, which is a non-cash item that we see in our COGS, spreads that out evenly. Importantly, the cost of all the labor, the cost of our management infrastructure, the cost of the overall running of the site gets spread now across a much higher volume level. The combination of those two supports us continue to be very bullish on the long-term gross margin trajectory.
We've talked about the automotive business being north of 20%. We continue to be very confident in that. What we're also beginning to see, which is an important point to note, two things, is the structural advantages that come from this very high degree of vertical integration, which in the beginning, at lower volumes, are actually quite painful because you have a lot of fixed costs that do not get absorbed across a lot of volume. As the volume starts to come up, we reach some flexion point where suddenly these things that we're building in-house and developed in-house become a real profit driver. That is one. In a similar way, there are a bunch of aspects of our business that we have built in-house that do not exist in traditional car companies outside of Tesla. That is essentially everything that exists downstream of the plant.
Distribution, sales, service, these are all heavy builds in the beginning to build out the infrastructure. On the distribution and sales side, our cost to deliver and sell a vehicle is far lower than using a third party, so paying a third party to do it for us. We will start to see the difference of us costing a couple of percent to sell and distribute a vehicle versus paying a third party, let's say, 10-15%. That will start to show up. Secondly, we will start to see service become a real contributor to our profitability. Of course, I think it is well understood the service side of the automotive business is very profitable. This is where a lot of dealerships make the vast majority of their profit.
In all the same ways, this is going to be a big source of profitability for us. We've seen that happen with Tesla, of course. We're going to start to see the same. In the beginning, it's all OpEx. As slowly as it starts to be revenue-generating, it transitions to COGS. Of course, it's highly profitable COGS.
Maybe on that, on the point of the R2, you've talked about a pretty substantial reduction in material cost, a 45% reduction versus R1. I wanted to touch on a couple of areas. One is the battery, and the other one is the electronic architecture. It's on the battery that maybe talk a little bit about the decision to transition to a 4695 cell structure. I think you've been using 2180 previously. I guess how you're thinking about the ramp curve for that cell. I mean, we've seen Tesla, for example, building their own 4680 cells, and they have made progress, but it's taken a little bit of time.
Yeah. Just to repeat something you said, just to make sure it's very clear. The R2 bill of materials is about half that of the R1 bill of materials. That's our current state R1, meaning with all these cost reductions we've made, we've still been able to take roughly half of the cost out again to deliver on a similar, call it as much as you can like for like R2 to R1 in terms of content and range. In terms of non-bill of material COGS, it's well under half of what R1 is. The cost structure on R2 is just meaningfully more advantageous than what we saw on R1. A big component of that, which I've referenced a few times here, and I just want to reiterate this, is supplier leverage.
When we negotiated the R1 BOM the first time, this was in 2018, 2019. We had very little. I mean, it was unclear who Rivian was. We had no customers. We'd just shown the product at the auto show, but that was really it. There was a lot of risk embedded in the price that we paid to our initial suppliers. We anticipated being able to negotiate that down following launch and the success of the products being pulled from so much demand for our products. With COVID and the supply chain crisis that hit, it just made that process much more elongated than we anticipated. In contrast, if we look at R2, the amount of leverage we have in these negotiations is night and day in every way.
We've seen R1 be highly successful in the U.S. market, where it's, as I said before, in terms of premium products, one of the best-selling products out there, the R1S being the best-selling premium SUV. The suppliers recognize that, and we've seen that built into pricing. Of course, it was really helpful that we signed a $5.8 billion software licensing deal as part of our joint venture with Volkswagen Group. That was a huge validation of our technology and, of course, a massive vote of confidence from the second largest car manufacturer in the world. That all led to a bill of materials that's really advantageous relative to R1. Now, saying that, there's also a bunch of design changes that have been integrated into the very core of how the vehicle is built. You asked about the electrical architecture and battery.
On the electrical architecture, we've taken what we did on the Gen 2 of R1, and we continue to push that further. In fact, in our last earnings letter, we put a few pictures of some of the computers, the PCBAs between what we call Gen 1, Gen 2, and then R2. Just visually, you can see they're getting smaller and smaller and smaller. We're getting far more efficient in how we design those. We're putting more of the content onto the chip itself. That's taking a lot of cost out. We're seeing that across every ECU and every part of the vehicle. Of course, the Volkswagen partnership actually helps that as well. In terms of the batteries, you said we're using a cylindrical cell, which is much larger diameter. It's a 46-millimeter diameter relative to the 21 that we're using in R1.
It is a taller cell. We are using a 95-millimeter tall cell. It is a 4695 instead of the 70-millimeter tall cell that we use with the R1 vehicle. That larger diameter, larger height, or taller height means that there is more energy in each cell, which means we need to really be thoughtful around how we manage those cells from a safety point of view. The cylindrical form factor is actually very safe in that it helps contain this large amount of energy quite effectively. We have designed a cooling system that really works around that. Importantly, it massively reduces the number of joints. The number of cells that need to be attached in the system, our Max Pack of R1 has 7,776 cells. Depending on the configuration, R2 has, depending on battery pack size, is low hundreds of cells.
That's a simplification. It allows the modules to be simplified. It allows the pack to be simplified. We've really done a lot of work on the high-voltage architecture within the pack. We've integrated the DCDC, the DC-AC, the AC-DC. All of that's been integrated into one housing that sits attached to the battery pack, which limits the amount of cable runs. I'm going to probably put a picture of this up on social sometimes because it's just so cool. It's like this beautiful integration of all these parts that used to be separate boxes into one housing. It's a large high-pressure die-casting housing that puts all that together. I mean, it takes a huge amount of cost out of the high-voltage architecture relative to what we have in R1 today.
Okay. That's helpful. Maybe this actually can kind of dovetail on that. Just all of the changes you've made, obviously, you're bringing down the price point as well with R2. That's going to open up a bigger TAM. How do you think about the, how do you think about kind of scaling production at the Normal plant? Also, is there a risk that Normal could reach a factory constraint before your Georgia plant comes online, which will also carry R2 as well as R3?
Yeah. We hope that Normal reaches a production constraint as soon as possible. That is the goal. We do believe that is going to happen, which is why we have talked about this, the importance of Georgia and bringing on additional capacity above and beyond what Normal can provide. Just for everyone on the line, Normal has R1 in it today. It has our commercial van in it, and we are adding R2. The combination of those three will have a capacity of just over 200,000 units a year. We think of it like a nominal capacity without really stretching into heavy overtime of 215,000 units a year. We are going to try to push that as much as we can to get beyond that. Of that, 155,000 are R2.
These are opportunities for us to push harder, but that's what its designed capacity is. The reason we've planned a ramp of this that's, I'd say, faster than what we did in R1, but very intentionally goes from starting with a single shift and then starting in 2026, early 2026, with a single shift and then adding a second shift in the latter part of 2026. The reason for that is the learnings we had around the importance of scaling the supply chain and scaling it in a thoughtful and predictable way. Some of the biggest pains we had in 2023 and even into the beginnings of 2024 were just the supply chain not all ramping at the same rate. We often think about that as if you're short a part, you can't make the vehicle. That is a challenge. That's a downside.
That's a problem with the supply chain that's not ramping consistently as fast as you want. The other problem is you have a bunch of the other parts that are still coming. You have to consume all these parts in this inventory while you wait for the constrained part to catch up. It's very challenging when you think about thousands of parts. If you think about discrete components, not source parts, it's tens of thousands of parts where any single one of those can stop production. Across hundreds of suppliers, it's a real orchestra of activity. With R1, we were learning for the first time when we launched the R1T, R1S, and EDV. We launched them all really in close proximity. Ingesting that much complexity was very difficult.
With R2, we're launching a very narrow set of build combinations, just R2. We're pulling in R3 after we've fully ramped. We are planning our supply chain and working with them very hard right now to make sure they're ready for this one shift to then two shift step that will happen in the latter part of 2026. We need it to be a very smooth launch, and we're busting our tails to make sure it's as smooth as possible.
Yeah, that makes sense. I wanted to maybe touch on the EDV and the commercial van business. You've got the partnership with Amazon. I think you've delivered 20,000 vehicles so far. The order was 400,000, so there's still more room to fill that out. You do have engagements with other fleet operators that you've spoken about. I believe capacity is for 65,000, but that can be flexed down for R2. How do you think about the addressable markets now and the mid to long-term opportunity there?
Yeah, Emmanuel, as you suggested, we're planning to adjust down the total output of EDV. If you think about the Normal plant, R2 will represent the vast majority of production with 155,000 of the 215. The remaining capacity that's in the plant, what we'll use that for is a combination of R1 and the commercial van, but most of that will go towards R1. That is really reflecting the commercial vehicle space and the commercial van space electrifying slower than we anticipated. I think some of the macro headwinds that we're seeing, both in terms of policy and I think also in terms of sentiment, mean that that space isn't electrifying as fast as what we thought it would do. As you call out, I mean, Amazon is a great anchor customer for us.
Because of them as an anchor customer, we were the top selling by volume electric van in the United States in 2024, which just demonstrates that the market is we were the highest selling, and we were not selling 100,000 vans. We were selling a much smaller number than that. It reflects just this space, even with us being the market share leader, it is just electrifying more slowly than we had originally thought and hoped. It is an amazing product, and we are starting to see more customers outside of Amazon start to use it. We are seeing some of the bigger fleets be more cautious in their own electrification plans relative to what they thought they would have done, let's say, a year ago.
Maybe just switching over to software and services, maybe we could just start with the Volkswagen JV. We get this question a lot in terms of the strategic perspective of the deal. I guess when we think about Rivian, and I think you talked about this a little bit earlier, one of the strongest attributes that you have is the ECU architecture that you've built almost entirely in-house. You own the entire software stack, including the base operating system and middleware. You've designed your own zonal controller. Certainly, these are things that legacy OEMs have not done and in some cases may not have the ability to do. We're working with Volkswagen where you'll be co-developing an architecture that actually now will be going into their vehicles, I believe, in 2027. They're also competing with you in North America via the Scout brand.
Maybe just talk about the benefits that you see of this joint venture relationship beyond the capital injection part of it, which I think we all get. Also, how you think about Rivian's ability to maintain a competitive advantage in software and network architectures.
Yeah, the partnership that we've built with Volkswagen Group has been, I referenced earlier. It's been, of course, focused, and the deal is, of course, designed around our network architecture. So it's our zonal controllers, what we think of as our experience management module, but essentially what runs the infotainment platform, and then the various layers of software that exist on those platforms. Being able to deploy that at scale across many different price points, different vehicle form factors, and across many different markets, of course, outside the United States as well, is great for us because it sees our technology driving impact in other markets. It is very mission-aligned to do that. What it also does is think of it as a front door into a really deep relationship where there's other benefits that come.
Of course, there's the efficiencies from a supply chain point of view of having a lot more volume on some of those shared components. You can imagine some of the SoCs that we're using across those compute platforms having volumes that are much, much larger. The volumes of building the PCBAs, the printed circuit board assemblies at the contract manufacturers now have a lot more volume. Extending beyond that into the items that are driven by that, some of the electromechanical systems that come out of the electrical hardware, there's now likelihood of more alignment of sourcing and therefore more volume. We see it as a really helpful corporate partnership that we have that extends and builds beyond just the joint venture. Volkswagen also has equity ownership in Rivian now as well as part of this deal. That's helpful.
It's really helpful. I said it before, and I'll say it again. I think far too often we look at the market side of this and we, it's not exactly why this keeps happening, but we sort of assume that there's a one-winner or one-player takeall. I think maybe it's because we've been trained in tech, in traditional tech, that that's how it works. There's going to be a dominant search engine platform. There's going to be a dominant e-commerce platform. In transportation, we just really don't believe there's going to be a single dominant vehicle brand. If there was, it would reflect a very significant shift in consumer behavior from what we've seen for the last 100 years. Meaning in the United States market, there's 300 different choices of ICE vehicles in terms of nameplates.
We think that level of product diversity is going to be necessary to go from 8% electrification to 100%. To date, it's largely been Tesla, as I talked about the lower price points. There have been other products, but they haven't been particularly compelling. To the extent that our technology helps Volkswagen Group make highly compelling products, which ultimately will give customers choice, we think that's great. We think it's going to be helpful to drive the overall shift in electrification. We think this is a very clear example of a rising tide will help all players in the space. As it pertains to Scout, we're, of course, part of that program, and that's part of our work with the joint venture. Does that feel like it's close to Rivian? Maybe.
I think the reality is the products are being designed and differentiated and branded and positioned differently than Rivian products. I think customers need that choice. I think it's going to be a really good thing for customers in the end.
That's helpful. Maybe we just have a minute left here, but I was curious about your opportunity in software and services across your own fleet, right? I think about this year, you're talking about $1 billion of revenue in software and services that you can generate. About half of that, I believe, is coming from your own revenue you're going to generate through service and software across your fleet. How do you see that kind of scaling over the next few years? Do you think ADAS could play a big role in generating incremental software and subscription revenue for you?
I love the question. We just, coincidentally, about a week, maybe two weeks ago, launched what we call our hands-free mode. Our Gen 2 vehicles have a massive shift or upgrade in the hardware on the vehicles. We went and vertically integrated our camera stack. We have a much more powerful compute stack. The Gen 2 vehicles have 55 megapixels of cameras, which is more than any other vehicle in North America. We have five radars, including a front imaging radar and then four corner radars. We have a really robust compute stack, which is powered by NVIDIA in terms of the in-vehicle inference. The headroom that that gives us in terms of capability is just remarkable.
The first sampling of that, and I use the word sampling intentionally because it by no means represents the end state. It is really the beginning, is a hands-free capability where you get on the highway, and previously you'd have your hands on the wheel. You could take them off for 15 seconds, but you'd have to put them back on. You can now have your hands off the wheel for the whole ride. That feature set, that capability will grow from a hands-free highway to increasingly approaching turn-by-turn so you can navigate from your house to the final destination in a hands-free, eyes-on environment. Next year, in 2026, we'll start to go hands-free, turn-by-turn, plus hands-free, eyes-off highway, which is a true level three highway, which means, of course, your hands come off the wheel, but your eyes can also go anywhere.
They can be on a phone. They can be in a conversation. They can be reading a book. The speed at which this is developing is just remarkable. The big shift, and I'm surprised this does not get more focused attention, is the way that self-driving was developed up until really only about two or three years ago was it was very rules-based, and it was very sequential, meaning you had cameras that identified objects, classified them, associated vectors with those objects, that collection of objects that went into a planner, and the planner then made decisions that were all rules-based around what to do. Those decisions would then get translated into a control strategy that the vehicle would then operate following the direction of the planner. What has changed so much is we now use AI to build a much more robust large-parameter model.
Think of it as a foundation model for the real world. We build that with lots of data coming off the vehicles. That large-parameter model gets distilled down into a smaller-parameter model, something that can be handled by the in-vehicle inference. The vehicle is trained using that with this end-to-end process. We no longer have to make all these individual steps in between. The speed at which it develops, the capabilities develop, is just remarkable. It is the same shift that we saw in the LLM world, where we've had voice assistants for years, but they were always a little bit clumsy. We started training end-to-end. We used modern transformer techniques. We thought about how we build these systems from a neural net point of view, and it was like a light switch went off. We think the same is happening in self-driving.
Saying all that, the big question, which is your question, is what can be charged for this? I think in the long, long term, who knows? It could be built into the vehicle price. We think it's going to be table stakes. If we say in the 2030s, I think it becomes table stakes that vehicles can do hands-free, eyes-off, turn-by-turn. In the next five years, I do think there's an incremental ability to charge significantly for these features where customers are willing to pay because it is creating real value for them. I think a few companies will have differentiated capabilities here. Of course, we believe we're one of those. Because it will be a subset of the full breadth of product offerings out there, we think there'll be this differentiated ability to monetize that capability.
We have talked about this with our customers. We said, "Look, today, our Driver+ feature is included. We will start to charge for it, and we are going to wait until the capabilities are so strong that it earns the right to have customers pay additional money for us to deliver that feature.
Okay, great. I think we're just at time. I think, yeah, pass it over to Emmanuel.
Yeah, RJ, thank you so much for joining us. We really appreciate your time and insights. Thank you, everybody, for joining.
Thank you, guys. Enjoyed it.