Carvana Co. (CVNA) Earnings Call Transcript & Summary

Hi, everyone. Thanks for joining. My name is David Lantz, and I'm a part of the retail hardlines team here at Wells Fargo, and we're very pleased to be joined by the Co-Founder and CEO of Carvana, Ernie Garcia, as well as Mike McKeever and Austin Knutson from the Capital Markets and Investor Relations team. Thank you guys for joining. And...

With that, we'll jump right into Q&A. So units have grown significantly for the better part of 2 years. What have you and the team done to get the business back into a growth mode?

Yes. So I would say our version of the story would be that I think we were -- we launched the company in 2013, sold our first car in January 2013. And I think we were in growth mode from then until early '22. And then I think we massively made a huge strategy shift where we needed to no longer be reliant on capital markets to run business and that led to us doing all sorts of things that caused us to pull back quite a bit on growth and actually go the other way for 1.5 years, 2 years. And then I think we just kind of stopped doing the things that were designed to rapidly improve unit economics and led to business shrinking a bit and move back to the same equilibrium that I think we had for the many years prior to '22, and we're now growing again. And I think the fundamental driver, I think, is just that we deliver experience that is desirable to customers that's not completely understood. And so I think we have to keep delivering great experiences over time. And I think as we do that, more customers tell their friends and that drives more growth. And then the business itself has positive feedback through inventory growth and everything else. And I think we just are leaning back into those levers.

Got it. That's helpful. And so you're aiming to sell 3 million retail vehicles a year in 5 to 10 years from now. So can you talk about the factors that could help you reach this in the earlier part of that or the later part?

Sure. So for perspective, for those of you that aren't that close to Carvana, we -- over the last couple of quarters, we've been run rating around 600,000 units per year, give or take. So I think getting to 3 million is around a 5x thing of the business. We've moved through several orders of magnitude since launching at 0 in 2013. And so -- and we're doing it into a market that has 40 million transactions per year. So that goal of 3 million units, while very large in absolute terms is only 7.5% of the market that we're in. So we think it's extremely achievable in terms of just the size of the market that we're moving into. And our business is relative to most growth businesses. I think it's very operationally intensive. We buy cars, mostly from customers, but also from auction. We put about $1,000 of parts and labor into every car that we buy. We photograph the car. It goes up on a website. A customer goes through a process to get approved for financing. We do verifications. They select their delivery time. We deliver it to them. We do title and registration. So there's many components of the business that are complex. And so our view is just continuing to scale that system at the rate that we're continuing to benefit from constantly growing demand for our offering is the hardest thing to do, and to hit that goal of 3 million units in 5 years equates to approximately 40% compounded growth, to do it in 10 equates to approximately 20% compounded growth. We think 40% is achievable, but obviously hard. And I think with relatively limited, if any, precedent for other businesses in history that have been disruptive concepts. So I think that suggests that it's hard but we've got a plan and we're marching as quickly as we can, and we'll see where we end up in that time frame.

How do you think about the balance of appropriately ramping production lines while also taking advantage of incremental near-term share gain opportunities? And I know in Q3, you had mentioned that the market is flattish and some peers have had some mixed reads since then. So I was just curious how you think about balancing the appropriate long-term stance with share gains today?

Well, I think our -- given our belief that ops is the likely driver of our time line to achieving any given level of scale. I think over time, we've been approximately operationally constrained. It's kind of how quickly we've -- what has driven our growth over time. I think we generally want to try to make decisions that smooth out ops as best we can, and make it easier for us to grow at high rates versus for sustained periods. And so I think trying to continually open additional what we call lines. We have production lines inside of our facilities and then open additional facilities, which means that we have more places to open lines in the future, means that like the -- we reduced the actual hardest operational problem in our business. It is likely scaling inspection centers. And the work that happens to scale the business happens at the individual center level. And so by having access to more of these centers, which is more work today than just scaling in centers that we already have that have excess capacity. But by opening more of those centers, we basically make it. So in the future, we have an easier scaling problem because we have more locations to do that work. And so I think right now, we're focused on trying to lay out a multiyear plan to grow at high rates for a long time. And that means continually growing inspection centers. So we've been continually doing that.

Got it. And recognizing that it's a really fragmented category, where would you say that your share gains are primarily coming from today?

I think our view is that our share gains are very broad-based. I think an interesting -- I think an interesting result of the automotive retail market being so fragmented is that publicly reporting companies are probably on the order of 10% of the market. In most retail verticals, the largest players, 20% to 30%, and the next largest player is 15% to 20%. And so if you have 4 or 5 public companies, you probably have the majority of that vertical that's reporting, and I think investors get a lot of visibility to that. I think in our market because it is so fragmented, a relatively small portion of the market is public. And so I think oftentimes, the narrative that is dominant in our industry can be dominated by what a couple of companies are seeing or experiencing. I think as a general matter, given how fragmented it is, if we were taking share randomly from all other retailers, and we're currently 1.5% give or take of the market. We'd be taking about 1.5% from everyone. That would obviously be very small and would not be noticeable to others. I think there probably is more concentration than like a true random draw across all the different retailers. But when we try to look at all the data that we have where we can look at all the cars that are listed online and where customers are buying from and we can look at cars that we value that then customers sell to a different dealer, we can see where those show up online and what dealer ended up buying those cars. And then we try to say who are we competing with? It tends to look like a mosaic of the entire industry more so than any concentrated number of players. So I think our general view is that we just got to stay focused on our customers, keep delivering great experiences and we know one player in this industry probably has a massive and direct impact on any other player in this industry. And we think that likely goes both directions.

And when we think about the go-forward growth potential and the 20% to 40% CAGRs that you mentioned over the next 5 to 10 years, how important is the used car auto backdrop in achieving that?

So I think as a general matter, used car retail has averaged for 20 years, about 40 million units, and it's probably been as high as 43% and it's probably been as low as 34%, maybe. I think the trough 34% was like 2009 or '10 post the great financial crisis. So I think generally speaking, like in our most recent quarters, we've been growing at about 40%. That's a very high rate of growth compared to the variation that we've seen over 20 years in the entire auto industry. So I think, our view is if the auto industry moves up or down by 2% or 3%, probably the best first guess is that we'll move up or down by the same amount. But that ends up just not being a huge part of our story in light of our growth. And so we're much more focused on laying foundations for sustained growth over a long period of time. And then I think we'll benefit when the market grows a little bit, and it will hurt us a little bit when the market shrinks, but on average about 40 million units.

And then I was in Orlando last week at the IRC tour, and there was some commentary around from the team that you're comfortable with the rate you're growing at the 44%. So can you dive into that a little bit more to especially as you get kind of tougher compares here over the coming quarters and into '26?

Yes. Well, so I mean, I think -- again, I think growth happens at a location level. Like there's many different kinds of growth, but like some growth that is pure technology like functions that are pure technology, things like credit scoring for us. They scale incredibly well, right? Like you just turn on more servers, do more calc, the scale just kind of like shows up and it's very easy. The other side of that spectrum is reconditioning centers. And in between that, you have customer care and registration and verifications and you've got logistics. I think when we look at all those different groups, we say, okay, let's look at the groups that have the worst scaling properties because you have the most labor and movement to physical things per unit of growth. And let's look at what we've achieved in the past. When we look at how we grew in 2020 and 2021, and we look back at -- we averaged 10 to 12 inspection centers back then compared to where we are today, we're like, okay, we likely by adding the same capacity per center that we did in 2020, '21, we can likely hit the last year at 40% growth going from a little over $2 million to $3 million, right? If you compound out of that 40% the entire time. So we think that it's a problem of complexity that we have seen before and kind of executed against. But it's hard to sustain that over and over. And I think when the team says that we're comfortable with our current level of growth, you are at an inspection center talking to the team on the ground that is adding lines on the ground, that's where work actually happens. And so, so far, I think the team on the ground, these locations and the corporate team that's overseeing a lot of this and the people ops team that's helping us hire in all these different locations they've done a really good job, and I think our growth has felt comfortable. And I think there's reason to believe that we should be able to continue to grow at high rates. But of course, it's going to be hard. And so I think we'll work really hard to make that as smooth as we can.

Got it. And since we're at the TMT Summit, I'd obviously be remiss to not ask you about AI. So how is the company using it today? And how is it benefiting both internal processes as well as the customer?

So I find AI to be sort of hard question to answer because it's just -- it's everywhere. And then also I know how skeptically I look at other people when they start saying AI is everywhere, and they know it means higher multiples. And so I can only imagine everyone looking at us with the same skepticism. But I think it's in every part of the business. At this point, I think what's really beneficial about Carvana is we built our business in a way that led to everything being kind of designed in systems that are API accessible and all decisions being deterministic and all those things being able to be calculated very, very quickly. And all of the data being stored in places that we immediately have access to because it's all first party, we're vertically integrated throughout. And so by virtue of just building the business that we've built over the last 12 or 13 years, we laid foundations incredibly well for an AI world where you can then take all of those services and you can take all that structured data and as long as it exists in a good place and it's immediately accessible, we can push it to these brains that can provide very high-quality answers very quickly. And so that shows up in obvious areas like our chatbot, it shows up in different tools we're building on the site today. It shows up in productivity throughout the entire organization. It's amazing what many of our different groups are doing. I mean our modeling groups have built basically like collections of agents at different levels where you've got an agent that takes in a data set and goes through and cleans it. You've got another agent that converts that into variables that likely have intuitive meaning. And then you have an agent that goes through and generates models, and you have an agent that goes through and skeptically evaluates the result of those models to try to figure out where there might be holes. And they're just now able to generate models that literally 100x the speed that they could even a couple of years ago. And then, of course, in like our development teams, development is like one of these areas where, I think if you ask 5 developers the best way to measure developer productivity, you're going to get 13 answers. And so I think trying to figure out what's the best metric to evaluate the speed at which we're increasing our output through our engineering teams isn't trivial, but I think we're materially up over the last 6 months by any of the metrics, and it seems like that's just continuing to happen. And so I think productivity, customer-facing capabilities, it's showing up everywhere.

Got it. And then shifting gears here to GPU. So you regularly talk about opportunities remaining across every line item within GPU. So can you talk about that in a bit more detail across retail, wholesale and other?

Sure. There's a lot in there. But I mean, so I think let's maybe -- let's move into retail a bit because we can maybe talk about it a little bit more deeply. But so I think in retail, there's a couple of things that are interesting that are going on. So I think that there's some like foundation laying that we think will lead to fundamental gains that will all else constant be a tailwind to retail GPU, which are things like rolling out ADESA Clear and having -- rolling out mega sites and having both retail and wholesale capabilities at all these sites so that we can be a more efficient buyer of cars. So basically, just between us and natural sellers of large pools of cars, we can cut cost out of the system and split the gains between us and them. So that's kind of like a structural improvement to the business that saves time and cost and energy and therefore, leads to gains. Then I think that we're constantly improving all of the ways that we buy cars and we price cars and we merchandise cars are informed by the quality of information we have about every car and the density of the information that allows us to understand how much any given feature is worth to a selling customer to a buying customer and where we're supposed to merchandise that. And that data, the sum of all that data that we have is growing at approximately the rate that the retail business grows. And so your -- for any given model, you run into noise at a certain depth. And if you're growing at 40% every year, you can move down like a level of depth roughly every year in terms of what you truly understand that you're able to price intelligently without being overwhelmed by noise. And so just getting smarter with -- getting more data and then being able to utilize all that data to price on both sides of the market, merchandise in the middle and build search tools in the middle as well, also helps retail GPU. So those are maybe like one is a structural advantage and one is kind of like a more of an analytical and data type projects. Both of those we expect to lead to increases in retail GPU, all else constant. And I say all else constant because we also definitely plan to pass back fundamental gains to customers as we think we're now at a spot where it's very easy to have line of sight to 13.5% EBITDA margins, which is a target that we set for ourselves for reasons we can explain if interested. And we think that there's enough fundamental gains to push us beyond that. And so we think it's going to be smart for us to utilize those incremental gains to pass back to customers to further differentiate our offering to enable us to play a really outsized role in this industry over a long period of time. So these are some examples.

Got it. And how do you think about the importance of passing fundamental GPU gains back to consumers as a function of achieving the 3 million retail unit target over the next 5 to 10 years?

So what I would say there is, I think -- I think a good approximation is that the quality of customer offering that we're giving to customers today versus the offering that we gave 10 or 12 years ago is about the same. And so we've driven for all about 1.5 years of that period, we've driven very outsized growth at constantly increasing scales from a base of 0 to a base of 150,000 a quarter with an economic offering that has been approximately constant. And so I think there's -- and at that level, you can look at our older cohorts, and you can see that we have higher market shares in those markets. And you can look at our younger cohorts and see they're ramping at a similar faster pace than the older cohorts were, so you can extrapolate out to much larger scale than we are today. And then you can look at those older cohorts, you can say they're still growing at very fast rates that approximate the rate at which the company is growing. And so you don't know exactly where that goes, but it suggests there's like a lot of headroom there. And then I think we can look at passing economic benefit back to customers and we AB test all that through rates and do customer bids on buying cars from customers and through pricing. And we have an understanding what those elasticities are. But to me, those are all just tools in our arsenal. And I think in the best version of the story, and I think something that's consistent with historical data, it's not obvious that you need to be passing back a ton of those economic gains to drive continued high levels of growth because I think that we've got lots of visibility to continue high levels of growth with our offering exactly where it is today. I think that's just kind of additional fuel that we think is going to be smart to use over time.

Got it. And on the EBITDA margin front, so you're in low double-digit territory today. Can you help us think through why 13.5% is the right target over the next 5 to 10 years? And I know you mentioned there's potential upside from that. But just as kind of a starting point?

Sure. So we can talk about for a long time, too. But what I would say is I think right now, like one way to articulate 13.5% is where we've been for the last couple of quarters, plus if you just look at our overhead expenses, and you assume that we lever that even partially with respect to growth on our way to these higher targets. It's very likely we would move through 13.5%. And then we're separately talking about all these fundamental gain opportunities that we think we have in every revenue line item and every expense line item that are going to be hard to unlock but that we plan to unlock over time. And then that suggests, okay, well, why don't use go well beyond that? And especially if you believe that with a fixed economic offering quality to customers. We've driven all the growth that we've seen since inception to today. And I think that's a reasonable argument. But then I think the other argument is that there is elasticity, and even in our 3 million goal, we would be 7.5% market share. And I think while it's important to have a goal at any point in time, it's not hard to imagine that when we get to that spot, we will look further down the field and have bigger goals. It's likely smart for us to pass some of those gains back because we can evaluate like what is elasticity and what is the value of incremental sale relative to the cost of getting that sale by handing economics back to customers. And that math suggests that we should be handing some back to customers and that we could take a really meaningful share. And I think as we do that, we're also competing with -- we're about 1.5% market share. For the most part, the other 98.5% market share of the market shares a cost structure and shares a revenue model that is very similar. If we put pressure on that, that business -- some of those businesses are not super well positioned to respond super easily because they just have a different lease structured business. And so it's probably smart for us to do it for that reason as well. And so I think 13.5% is not a perfect scientific exercise. It's like a bunch of science goes into it to approximate and then you try to pick a reasonable goal, and that's the goal we picked.

Got it. And on ADESA, you're completing about 10 integrations a year. So curious how you're measuring success there?

So I think the success is measured by opening the sites, doing the technology conversion so that we have the ability to run our play, getting leadership in place at those sites and then beginning to scale them, produce cars and then have those cars -- we compare those cars and we do it through what we call a global local process when we look at any given facility that's doing any given function, and we compare how that facility is performing compared to the other facilities that perform the same function. And so in recon, we want to look at cost speed and quality and just say, how well is this location doing in those 3 dimensions relative to our other locations? And how quickly can we climb that curve to be as efficient as other locations. And I think the steepness of those curves is a major way that we measure it.

And as inventory pools ramp, can you walk through some of the benefits and risks associated with that?

I think -- let's start with the risks. The risks are it's inherently harder to open a new location and ramp up from 0 because you have the new technology rollout, you have new leadership. You have all new people you have to hire and train that don't get to learn from preexisting people that are already there doing the exact same thing. And so it's just kind of like inherently a harder problem. And I think that when you're taking on any operational problem, the harder it is, the higher the likelihood is that you stumble. So I think that's the risk. And then I think the benefit is very mechanical. It's just in the most efficient version of the Carvana machine, you want to have many locations around the country that are distributed approximately similarly to the way the population is distributed so you can buy cars and minimize the transport to the location we're going to do the reconditioning and then merchandise it and then have ideally as much density in each one of those locations, so that customer as much density you can with the best distribution of inventory so that customers nearby are most likely select cars that are nearby. So you also have less outbound transport. And so just the more locations that we add necessarily the shorter the distances for the average inbound transport and then either the short of the expected distances for the average outbound transport or the more selection you have that's available faster depending on how you want to position the system from like a shipping fee perspective.

Got it. That's helpful. And so there are a lot of concerns out there today around the credit environment. So curious if you could talk about how you see the backdrop today?

So I think you always look smarter to be cynical and skeptical, and we have a really strong desire, especially in this beautiful stage to look smart. So I want to be cynical and skeptical, but I also think the data that we look at looks pretty good. And so I think it's been interesting to try to do some work on our side to tie out why do a bunch of smart investors feel like they're really anxious about credit. And why do most originators and investors in credit feel like things are kind of moving forward about as expected. And I don't want to say that we have all the answers there, but I do think there's a good chance that a decent part of that puzzle is that every originator, I think, had tough '22 and '23 vintages and then virtually every originator tightened credit in late '23 or early '24. And I think if you're looking at portfolio level metrics, I think the '22 and '23 vintages are playing a bigger and bigger role in the overall portfolio and causing things to keep -- look like they're getting worse. But if you're looking at actual vintages, I think it was easy to forecast where those were going to be 12 months ago, and most of the '24 and '25 vintages look pretty good. So it seems like consumer credit is better than the average narrative out there, but maybe we end up being wrong on that over time.

And then there's a big opportunity with brand awareness, obviously, with where your market share sits today. So can you talk about how you're going about capturing that and how you're using advertising as a lever?

Sure. And so I mean, to me, I would just say that like I think one of the interesting things for any like analytical mind to look at is to look at the cohort curves that we gave out every year up until I believe 2021, and you can kind of see that our oldest cohorts are ramping at a certain speed and then the newer cohorts were all ramping at a somewhat similar speed. And there's this question like, okay, if you launched the market and then a bunch of time passed and you got to a certain market share, how come when you launched that next market, why didn't it just start at the same market share that the other market was already at? If you have inventory that can be shipped around the country and you have prices that are the same, you're bidding the same for customer cars and you've got financing that's the same and you roughly are using mostly national advertising channels. Why don't you start at that same point and just kind of instantly jump there. What we actually saw was you started the origin and you built up over time. And to me, I think it's because in many customer conversations, it becomes clear that the actual question customers are asking themselves is they're saying, Carvana seems appealing. Like, first of all, I don't know what Carvana is. It's -- I've seen some ads and it's that vending machine company and how does it work? That's how most customers think about Carvana. Then they learn more. And it's like, okay, not going to a dealership for 4 hours sounds appealing, a broad selection sounds appealing, a reasonable price sounds appealing, a 7-year-term policy that sounds appealing. But the other thing is I'm anxious about buying a car, right, like buying a car is an anxiety producing customer experience that has a reputation that has been earned over a very long period of time. And so part of what I'm trying to do is have it be better. Part of what I'm trying to do is make sure that I don't do anything dumb. And a lot of times, the way to feel the least dumb is to do the traditional thing. And so I think many customers are like that's the debate that they have in their head. It's like, okay, Carvana seems appealing. But also if I just buy a car, the old way that everyone has always bought a car, I can't be that wrong. And I only buy a car once every 5 or 6 years. So what do I do? And I think what brand advertising is about and what delivering good customer experience is about is like if you have something that is actually fundamentally better, you just need that story to be in the minds of consumers to overpower the anxiety of I just want to make sure I don't look dumb. And so I think that -- most importantly, that's about delivering good customer experience over and over again and having friends tell friends about it. But it's also about leaning into advertising a little bit and making sure that our brand is out there because even things like we all kind of know this. If you see an ad on Monday night football, you think that company is legit, right? You just do. Like that happens subconsciously, you believe that. And you don't necessarily believe it if you -- if it's -- or you'll believe it to a lesser degree, if it's a low budget ad during a local news broadcast, right? You're going to feel differently about those things. So just getting out there in the world and having people see you and make all those subconscious connections so that they don't run the risk of feeling like they made a mistake is I think a huge part of what we have to do over a long period of time to go from 1.5% market share to 7.5% to be on that.

Got it. And then on the same day and next day delivery, how are you measuring success of the pilot in Phoenix?

I think we feel like the benefit to faster delivery in conversion are very well understood. So I think it's about how well we're executing, how many customers we can get to have options where they have a same-day delivery option available to them. And then what is the customer uptake and then how well do we actually fulfill that promise, that's, I think, generally how we're measuring it. And I think the progress that we made there is very impressive, and it's more -- I think for a business to be able to grow really fast long period of time, you want it to be as simple as possible. But for a business to have moats that are defendable, you want to be as complex as possible. We happen to be the latter kind of business. It's complex. Same-day delivery is not like a make a strategic decision, hire more people, deal with a little bit less labor efficiency and you've got it. It's -- there's a lot to do because like you have to manage different complexity of finance verifications for different credit types. You have to manage car picking, you have to manage presale inspections. You have to manage title and registration that varies by location. You have a lot of things that you have to manage that require that we kind of reoptimize our systems and design for them. And then you have to make sure that you also have staffing so that you can fulfill that very quickly. And so it's like a real undertaking, and that's why we focused -- we rolled out the capability to many markets and sort of like our system as it was already kind of naturally designed without that specific intent and our staffing models as they were already designed about a specific intent. We let same-day delivery sort of happen in that way. And what's happening in Phoenix now is we're purposefully trying to increase the availability of that option to many more customers through system design and through staffing models. And I think we've seen a lot of success so far and we hope to continue.

And to what degree is a broader rollout of same-day delivery embedded within your assumptions for 3 million retail units a year?

I mean I would go back to, I think a lot of what we're doing when we're estimating future retail units is we're looking at market shares of our earlier cohorts that are higher and then we're looking at the growth rates of those, we're looking at the various elasticities across the business. We have estimates for when you grow inventory. What does that do to conversion, what does that do to scale? And how does that kind of feedback loop play out when you -- we've made delivery times continually faster over that entire period of time. We have estimates of the feedback of that. And so we can kind of look at those things and extrapolate them out. And they definitely play into our models and into our confidence that, that kind of a number is achievable, but like we haven't broken that out precisely.

And for those not familiar, can you talk about the differences between the full build-out of ADESA locations that will begin in 2026 relative to the integrations that are already ongoing?

Sure. So -- the biggest difference is that auction locations forever, like as a seller of a car, when you go to an auction, you have an option to say, replace this windshield and clean up this headlight and bang out this dent over here. And I'm going to spend $500 to do that. And my hope is that I'm going to sell the car for $750 more. And so most auction businesses have some like reconditioning/like mechanic capability that was already built into the auction business itself. So there's building and space for that. So a lot of the initial rollout that we were doing was utilizing existing space that was underutilized maybe putting in some lift, putting in our software, putting in management teams and leadership in hiring and training and then pushing reconditioning through that pre-existing footprint. The full build-out means now we go to a site that's 50 or 60 acres, where maybe the initial build-out that already exists as part of the auction was enough to support 1 or 2 lines. And we want to have building capacity to support 6 to 8 lines. We then need to maybe build another building, put down a photo booth, put in more lifts. And so that's more capital intensive. And so that's why I think we've been able to open some of these locations with very low capital intensity to sort of get a toehold for these different inventory pools and leadership and everything else. And then I think we will now go through and start to fully build out some of these facilities where we can do meaningful volume out of each one.

Got it. And then can you walk through some of the differences and to your point, the capital intensity of the fuller build-outs versus the integrations that are ongoing and kind of split that out?

Yes. Well, the integrations that we've done so far, for the most part, have been very low CapEx. And then when we bought ADESA, we sized the CapEx that would be required to do the sum of build-outs to get us to 3 million total reconditioning capacity to be around $1 billion. And I think that remains like a good ballpark estimate. There's been some inflation since then. I think it's probably ticked up a bit, but that's a good first order estimate. And so that's -- you take that divided by the number of sites that we'll do, and that's approximately the investment that we expect at these locations.

Got it. And it looks like we only have a minute left. So one more here. Can you just talk about the significant -- the timing and the significance of the recent upsizing and new loan sales that were announced in Q3?

I think -- I mean I'm extremely biased here, but I think as a general matter, consumer credit is a deeply fundamental and highly valuable asset for people to invest in. And as a general matter, it's relatively hard for even pretty sophisticated investors to get access to many of these types of loans. And I think the way the markets have evolved there over a long period of time is that, generally speaking, you have companies that have kind of paired capital with sales teams that go out to many dealerships with verification capabilities and real-time credit scoring capabilities so that they can originate these loans and then investors had access to the equity in those companies. What our business does is we separate that. We basically have our own flow of customers that are coming to us. We've got our own credit scoring. We've got our own credit pricing. We connect it to verifications and the servicing. And then we create an asset that is easier for investors to invest and without having to build all these operational capabilities. And basically, we think that -- in many ways, that's the exact same trade that is happening in all of fintech. So I think that in automotive, we're sort of doing that same thing and that opens you up to an entirely new investor class. And so I think we've got existing investors that we have long-term relationships with that we're still selling lots of loans to and have upsized, and then I think that we've been gradually getting deeper relationships with other new investors that are realizing even as they held loans, they went through '22 and '23 that were not great vintages -- the downside cases look pretty decent. And so they're more excited to invest more money in those assets. And our hope is that as we continue to get more scale, we'll be able to do that more. And if anything, drive down our cost of funds and therefore drive up our finance GPU, but we have to execute that to happen.

This has been super helpful. Thank you, Ernie.

Well, thank you. Thanks, everyone.