Upstart Holdings, Inc. (UPST) Earnings Call Transcript & Summary

All right. We're going to go ahead and get started. I think, according to my watch, it's 20 seconds early, but if we start early we end early, so we'll see how it goes. Anyway, thanks for joining us everybody this morning. I'm James Faucette, senior research analyst at Morgan Stanley, covering fintech. Very pleased today to have Dave Girouard, CEO of Upstart with us to chat about his business. Before we get going, maybe I want to avoid getting put in jail. So I'm going to read my quick disclosure. For important disclosures, please see the Morgan Stanley research disclosure website at morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley representative. You got anything you need to say to stay out of jail?

No, just you said that.

All right. Good. Look, Dave, thank you so much for being here. We were chatting earlier before we started. You guys have been committed to this conference for a while. And so hopefully, you're as excited as we are for you to be here. Maybe I'll start. It's been a tremendous year for your team, both on a stock performance perspective, and it's been a roller coaster for sure for a lot of people. But also, more importantly, at least from my perspective, from the fundamental results perspective, maybe we can just quickly kick off with an overview of Upstart, along with how you gave the long-term value add for the business and to both the underlying borrowers and your bank partners just kind of set the stage for us as to what Upstart is? What its aspirations are? And where you're driving?

Yes, sure. So we're about a 10-year-old company and really focused at the intersection of artificial intelligence and lending. And lending obviously, is enormous many trillion dollar industry. Banks, nonbank lenders, consumer lending, business lending, whatever. It's an enormous cornerstone of pretty much everything we do. It's also sort of individually important to people. And -- and in our view, it's an incredibly inefficient market that does not serve the borrower well and frankly, it doesn't serve the lender particularly well and artificial intelligence has the potential to change that pretty dramatically by having superior risk models and automation, et cetera, the kinds of things that AI does really well. Our business is unique in that. We aren't a bank. We're not a lender. We don't intend to be. We're a 2-sided business in the sense that we are a consumer brand. We do market to consumers. But we're also kind of a B2B company, providing AI technology to banks and credit unions who are our partners. And so kind of a interesting technology for sure, with a lot of sort of issues, questions that people need to understand. And then a business model that's quite unique that makes us, I think, from a public company a little bit difficult to sort of map us to anybody else. We don't look like anybody else certainly in the public markets that can be a challenge. But I will also say the greatest companies -- greatest technologies companies in the world are classified themselves. Nobody else looks like Google or Apple or Facebook. So we kind of think that it's okay to be unique in the long run.

Sure. And that's understandable. I guess maybe kind of diving into the current environment, et cetera, and how that fits with your business and your business model and structure. I'm going to kind of get right into the -- a little bit the nitty gritty to start because I think that's at the forefront of a lot of people's minds. And then we'll kind of look at the broader opportunities as we go through our conversation. But delinquency rates across multiple consumer products have started to normalize off of pandemic era lows, right? Like we had a lot of stimulus and other things that contributed to kind of people being able to pay at rates that were historically better than normal. Have you seen an impact on your business so far from this normalization process? Whether that's consumer demand, partner return requirements or other things that I'm sure you get asked a lot about.

Yes. So I mean, first of all, as a technology provider in a very cyclical industry, it's sort of fundamentally important that we are helpful when times get -- when things get jittery, when volatility is high, when the economy is changing. And that's part of our value proposition to our bank partners as well as to the consumers as we can done right, help lenders navigate turbulence in the economy and on the consumer side, continue to make credit available at times when it's needed the most. And that's one of the sort of cores of our business is lending traditionally been very human-based with a few simple tools, but largely intuition and human-based in terms of when to do how much of it, et cetera. But our view generally is unsurprisingly software and very sophisticated AI models can make better decisions than humans, particularly when it gets noisy out there. We're in one of those noisy times now as we have been for a couple of years now. The beginning of COVID was a bit crazy, unemployment going from like 4% to 14% in just a few weeks. As you said, we had a lot of government stimulus really pushed down default rates unnaturally, and probably in our world, 40% to 50% below what our models would have estimated. But we kind of took the opportunity to build the notion of stimulus -- government stimulus into our model. So you can think of it as our models are getting a little bit better over time at dealing with the macro. Now we're normalizing, stimulus has largely gone away, consumers are kind of returning to hold spend habits, et cetera, and default rates have sort of gone back roughly to 2019 level. And that's the job of our software is like to give banks and lenders the comfort that it will handle that right. It will do the right things. It won't price things too low because there's government stimulus driving down default rates, et cetera. And so yes, I would say generally, it hasn't -- the return has not negatively impacted our business in the sense of -- there's a lot of liquidity on our platform. Rates have moved around a little bit, but our platform is not terribly sensitive to interest rates, either the Fed rates or what how -- what the consumer prices that -- the prices the consumers are seeing. So hasn't directly affected though there's a lot of noise certainly between the potential for the Fed to raise rates, recession -- not recession, but let's just say, inflation and what impact that might have. So certainly, there's a lot of chatter, a lot of noise out there. And of course, the geopolitics of the Ukraine just has a lot of unknowns out there. So times like this, in our view, is they're normal. Before COVID, we were probably in the personal lending space, our platform is probably fourth or so, largest in the U.S. And by the time we got to COVID, we were by far #1. So I guess in our view, -- from a strategic perspective, volatility and turbulence is length to us. It is an environment that we shine in, and we think we'll continue to do so.

So you mentioned there to be the amount of liquidity on Upstart's platform, et cetera. An important component of that liquidity at least indirectly, has been the securitization markets for your partners, et cetera. There seem to be a little bit of spread widening on securitizations in 2021, and that seems to instill a lot of concern in investors. How does that relate, though, to demand within the institutional investor funding channel for Upstart. And what are the commonalities? Or what are the things that those securitizations are indicative of versus what aren't they indicative of?

Yes. It's useful to know how securitizations play in our platform. So first of all, when a consumer applies for a loan, there's 40-plus banks and credit unions, who essentially compete for that borrower. If the borrower has the right sort of credentials and risk level, et cetera, banks can control this completely themselves. They effectively compete in what is effectively an auction. The consumer will see the best offer available to them. Now for a variety of reasons, those banks don't make an offer to that person. There are a few what we call conduit banks, who are -- really have -- you might think of if you're a tech geek, you might think of it as a kind of virtual -- almost unlimited virtual balance sheet because behind them are 150-ish institutional buyers of credit that have much broader and diverse risk appetite than any bank could ever have. So think of it as that's how we get such liquidity is banks take what makes sense for them. They're using depository capital, so their offers are extremely competitive, but they don't have the risk appetite that the whole capital markets would have. But to get back to your point, none of that is immediately securitized. And in fact, any loan that's originated on our platform there's never -- we would not know at that time if it ever will. There's certainly no promise it will be securitized. We do sponsor securitizations periodically, typically once a quarter-ish, and that means all those institutions, those are credit funds. Those are SEC registered funds. They can be investment banks. They could be -- there's a wide-ranged insurance companies, pension funds, they can choose to securitize the -- we make a securitization platform available to them. And if the trade is great, and they want to do it and they're happy to get the cash back and have something else to put the cash into then they do it. In 2021, the execution of the ABS markets was extraordinary. So a large fraction of them were taking advantage of that and no reason they shouldn't. I mean if you can buy a loan at par and then sell it at 108 within a month or 2, who's not going to take that trade assuming you have something else productive to do with that capital. And that's the thing. It's really -- I think it's good because it's -- liquidity is always better than non-liquidity. ABS markets offer transparency. You can see how those loans perform. It's very public. So I think it sort of is a value add to the investment community that buys our product. And -- but it's not always available. We would certainly expect because ABS markets have certainly gone cool since 2021, that a far lower fraction of loans from our platform will end up in the ABS markets this year, I mean, that would be our prediction. But it's not -- it's almost a tech secondary or even tertiary step in the process.

So if the ABS markets cool, et cetera, what's your impression of how that impacts demand for your loans from the primary buyers of them?

Well, the majority of buyers don't securitize anyway. So of course, it has no effect on them. And then there's a bunch that will securitize if it's really productive to do so and profitable to do so. Otherwise, they'll just happily hold the loans and appreciate and enjoy the yield they get from them. There's probably a small fraction that would maybe have less of an appetite unless there was an ABS outlet at the other end. But I don't -- in our view, that's not all that material to us because the vast majority of them either aren't securitizing at all or we do it opportunistically, but don't have a need for it.

So I think -- I appreciate that. And I think that's -- I wanted to hit these kind of issues because that's the question that we get most often from investors that are first taking a look at Upstart is trying to understand kind of these dynamics and consumers, the consumer market and then the associated securitizations and what information may or may not be gleaned from that. But -- what I find really interesting about Upstart as a company and as an approach and as the strategy is that a lot of fintech, and I would say the majority of fintech is really to find by either the things that they're doing as companies that allow them to operate in a different regulatory regime than what we often see, the incumbents objected to or frankly, they're just operating in a way that they're like, look, we're not going to make as much money as an incumbent. But we think we can build the base and eventually will be profitable when we monetize that customer base. But to me, Upstart is true fintech. Like you're trying to apply new technology to financial services and the financial services sector. You highlighted it initially just the software and that kind of thing. You feel like can do a better job of underwriting than the personal-based underwriting that is often endemic to incumbent institutions. So let's start with though, how do you start to measure that versus more traditional underwriting methods like the performance of the loans? Or even if people are using FICO et cetera. How do you measure your models underwriting performance vis-a-vis those types of loans on a relative basis? And what are you trying to get to?

Sure. So of course, a bank looking considering working with us wants to understand, is this model really different? How do we know it's different. There's a lot of different ways you can measure it, none of which are perfect, but all of which point to the same thing that you can look at. For example, when you securitize loans, the rating agencies will put a loss estimate on a pool of loans. And then invariably, our losses end up being roughly half of what they estimate they'll be. And it's because they're using a traditional kind of FICO DTI-based approach to estimate what losses will be and suddenly. And this has been the case since the first Upstart securitization in 2017 is that our losses are always dramatically lower. And that's not the case across the board. It's not that they are putting a lot of buffer in there. If you actually look for most of the securitizations, their estimates are more or less roughly on target. So that's one way. Another way is we do back test with banks to see how they would -- how we would perform relative to how their model -- what decisions their models would have made. Sometimes we do shadow pricing. So essentially, we're just running silently in the background. And generally, what you can see is that we can double or even triple approval rates at similar loss rates. Or if you sort of turn the dials in a more conservative fashion, we can eliminate as much as 3/4 of the losses, keeping the approval rates the same. So it's just to us, none of us -- I did not come from lending or credit honestly. But to me, as a technology person, having spent all my years at Google, the idea that enormous amounts of data, very sophisticated models can more accurately predict the performance of a loan is like maybe obvious, right? But it's been so for many reasons, so difficult for this technology to come together in this industry that there still remains enormous disbelief that it could really be different. There's also been historically, if you go back 20, 30, 40 years, many companies who have come along and claimed a better mousetrap in credit. And then a recession hits or whatever, whether it's First Marblehead or it's that the first whatever the name of that credit card was first offered over the Internet. It's just this thing that was a little shiny for a while and then oh turns out it wasn't so great. So that puts our company in a place where we have kind of a belief adoption curve, what I would describe it as, which is it's just plenty of history that suggests what we're doing is either impossible or so hard that it can't possibly be true. But that's okay. I think we're proving it month by month, quarter by quarter. And ultimately, it comes out in the results of the company. There's not other fintech companies growing at triple-digit rates and actually very profitable. So clearly, that's happening somehow rather.

Right, right, right. And as a quick follow-up on that. So as we've gone through this macro change and so things are starting to normalize, et cetera. How quickly does your underwriting -- are you able to respond to those kinds of changes and incorporate that into the loans that you're putting together?

Yes. So the model can change and learn in multiple ways. The most obvious, the most fundamental because it is AI and machine learning is as performance changes, as delinquencies spike or something happens, it will tweak itself and learn. And I mean, that's the heart of machine learning is responding to the actual events in the real world that are relevant to them. So that's the most normal way. Because the system also has some macro awareness. It knows from the Fed, what credit card default rates are, it has a sort of consensus view of where unemployment is and where it is expected to go. It can sort of get ahead and do some things even before seeing the actual results. So I would say the former, which is actually responding to actual performance of loans is very sophisticated. That's what we've been doing most of our time. In the recent years, we've really been kind of beginning to upgrade how it just understands the macro environment it's operating in, that's still relatively unsophisticated, but the first part is quite sophisticated.

So here's like maybe not a philosophical question. It's just like a broader question as I like pull out and look at things from a higher level. And you talked about how you quickly become the #1 personal loan originator in the U.S., et cetera. Is there a point at which -- and what is the point at which your loan performance and that your underwriting starts to look like the market, right? Like is it just when you are the market or like what's that point from your perspective? Or is there a point?

Well, the starting point I would say is my co-founder likes to say, every loan that defaults should never have been originated and every loan that is paid back in full, the borrower paid too much. right? So that sort of suggests we're on a never-ending journey. If you could randomly offer loans to every American out there, by our best guestimate, 85% to 90% of those loans would be paid back. And so 10% or 15%-ish would default. So that sort of means, we are on this path to identify who those 85% to 90% of people are and give them effectively loans at risk-free rates. Now that's sort of we'll never quite get there. But that's what the market should look like. If the systems were perfect, which they never will be, that's what it would look like virtually risk-free rates to people who are guaranteed because of the model is so damn smart to pay them back. Now of course, we won't get there. We can't get there. But we're so far from there as an industry that taking big steps toward there creates enormous differentiation, and it's creating the success you're seeing at Upstart. And I will further say, if you're at all a sort of technology historian. If you care like where technology came from in the 70s and 80s, where it is today, and you know the banking and lending industry, it's implausible that 10 years from now, 20 years from now, virtually every loan in the world is going to have AI underneath it, making a smarter credit decision. It just seems incredible that it wouldn't be true. How much of that we can power and win? I mean, that's up to us to execute, but the sort of likelihood of almost all lending, in all corners of the globe becoming AI-enabled to us seems fairly obvious.

So I think that makes a lot of sense. And -- which is -- and like, I generally believe like analysis is better, et cetera. But I'm struck a little bit by there's comments from time to time from members of the Federal Reserve, et cetera, how they want to -- they think it's important preserver or even encourage personal-based underwriting for lending. And whether that's personal loans or more in the commercial space, I think there can be some nuance there. But -- like what do you make of like those kinds of statements, where, like I said, and these are well-informed important people within the financial system.

I haven't heard them. I'd just describe it as nostalgic. I mean there's no losers in a system where better risk models and more automation, make it work better for the lender and better for the borrower. There's no losers in that system. So in fact, we have double-digit improvements to credit access to every demographic in the U.S. We're entirely consumer-focused at the moment. So every gender, every race, ethnic origin, age, take your pick, we are dramatically improving credit access in the categories we're in today for every demographic. And when you get those facts out there, whoever it is that is still nostalgic for the old way of doing things, quickly goes wow. And one of the things I want to say is the access to credit in the U.S. has not changed materially in 30 or 40 years at least.

And in some ways, it may be worse for some people.

It may well be. I mean think of what else has changed. Access to information, access to broadband. -- almost so many parts of our lives have improved dramatically in the last 34 years. But credit access more or less hasn't moved. And I think we're demonstrating it is possible to move the needle quickly and materially for the American consumer. And so our hope is we're going to keep proving this until we get more and more regulators, lawmakers, banks, lenders behind this notion. And we don't remotely expect to be the only purveyor of this. If it's as big of a secular change as we believe it is, there's going to be plenty of others doing it. We think we have a heck of a head start in a pull position, but it fits as big as we believe it will be. I'd be shocked in 10, 20 years, there aren't 20 others around the globe, covering different parts of credit. And we're certainly going to participate as broadly as we're able to.

So I've been monopolizing questions. Any questions from the audience, just raise your hand, we'll get a mic to you. We've got the runners in the back. We've got one here in front. Well, they're bringing the mic up to you. I want to go back to the banks and the bank partners. It seems like you've gotten really and you still have really solid momentum behind additions of bank partners. Can you talk to us a little bit about the opportunity there? How much there still is to go in terms of adding bank and lending partners on to the platform? And what's happening both in terms of what you're seeing regarding the ramping product usage? But also in terms of growth trajectory on a multiyear basis, how big is the opportunity really at least from a TAM perspective?

Yes. So I mean, this 40-plus banks, credit unions and the partners on the platform today, that's a rounding error, frankly. I mean, there's probably 5,000, 6,000 banks in the U.S. equal number of credit unions, all of whom are lending -- a vast majority of whom are lending using an approach that's been around for decades. And so that's why it's important to have a consumer side to our business, too, because the matrix of banks and credit unions using Upstart today is quite sparse on a national basis, but we can very effectively direct consumers to the ones that are, and that's the best credit product available to them. So that's why in the early days, is very helpful to not just be a seller of technology to banks, but to also be able to sort of push the market towards the ones using our technology. But yes, we have a long way to go. I mean it's still early stage of adoption. And it's unsurprising. I mean, I came from Google and with starting selling Gmail to companies. And you can imagine in 2006, to say to a company, "Hey, retire those exchange servers, get rid of IBM, Lotus nodes if you can believe that. And just use Gmail, you could sort of get laughed out of the room. But of course, that's all changed completely today. And in the same way, I think banks will feel this way about using AI to do lending 10 years from now without question.

Question here?

How do you expect the recent market volatility, higher inflation rates and maybe some of the credit tailwinds that -- we might be at the tail end of that impact your business? And how do you think about the funding mix given higher interest rates and maybe impact on the ABS market down the line?

Yes. I mean I would just say it's really -- we are, by definition, we serve a cyclical industry, and it is our job to serve it particularly well when it's complicated and noisy out there as it is right now. And we're adept at it. We're -- maybe think of it in 1 or 2 ways. What's our value prop to our banks and lending partners? And then how do we, as a company ourselves survive it? So I think the value prop gets strengthened to banks. It helps them deal more adeptly with changing environments and such. So I don't think -- I think if anything, more noise, more volatility ends up being a good thing for us as a company because we can handle it, manage it better than others. Our own business, of course, we're a very lightweight company. We're a small balance sheet. We have low fixed costs. And honestly, the truth of Upstart is we serve a cyclical or a industry, you better be ready for volumes to go up and down, and we fully do. We don't assume our volumes are always going to go up into the right based on the nature of the industry we serve. But having said that, even during COVID, even the first few weeks of COVID, when almost every bank in the country just said, put the brakes on, let's stop. I mean Upstart didn't really lose significant money even in what I would consider to be probably the most extreme handful of weeks that we've seen since God knows when. So I think in the end, we're a company that thrives in this type of environment, and we will, without question, come out stronger on the other side of whatever it is.

Bobby?

Given the technology advantage you're bringing to the customer and the solution, how do you price the product relative to the other older solutions in the market?

Well, the short version is the core of our technology is predicting loss rates and prepayments on a monthly basis for a loan, okay? And that's the heart of pricing a loan and deciding whether a loan can be offered is what is the expectation of cash flows of that loan on a monthly basis, which is itself the most sophisticated part? The rest of it is really up to the bank and the lender to decide what is the return target, what's their max loss rate they're willing to accept, et cetera. And that part is just arithmetic, frankly, where it says, okay, here's the cash flows predicted for this applicant. The bank might say, okay, I need a 3.5% ROA after X, Y and Z. And there it is. So the price is entirely a function of the risk our system is assessing and the business objective, if you will of the lender put together. That's what the -- and that results in what the consumer will see in terms of an offer or not.

Sure. But the revenue to you as a function of the solution.

That's really straightforward. We're a fee-based business. So we -- almost all of our revenue comes in 3 forms. A referral fee, which is when the consumer finds us and we direct them to the bank, which is not always the case, that's a referral fee, which is roughly 3% of the value of the loan. And then a fixed dollar fee for the -- for actually pricing originating the whole process of originating the loan, a few hundred dollars. And then mostly, but not always, we service the loan on behalf of the lender and that's sort of a fee over the course. So those 3 fees add up to almost all of our revenue.

Thank you. Got it.

There's a question here.

Dave, just a quick question for you. Compared to your credit card rates, so you offer slightly lower rates than say, credit cards. How much lower do you usually offer?

It's not something we can obviously measure. I mean, I would say we've done studies in the past to say what our borrowers on our platform paying on credit card rates. Typically, they are people that are paying in the 20s, 22, 24, 26. And then what we can offer them is, again, a function of what I just described. It's not -- we're not like trying to go 2% under their credit card rate or anything else. We're trying to, based on the risk of the loan and the business requirements of the lender make the very best offer we can. The reason we do so well is because there's a quite -- generally a quite large spread there.

So say, if a credit card rate is 20%, what Upstart will offer may be a few percentages lower, maybe 17%, something like that?

I'm sure these people will get 19%. There's probably someone who gets none because we think their risk is actually higher than that 20%. So it's -- there's no singular answer to that question other than clearly, we're a good business because a significant fraction of them are seeing a lot of savings potential by moving from a revolving and credit card to an installment loan with an end date.

I see. Second question. So your sticker sauce, the magic sauce that your company has is basically the AI, the algorithm in loan origination. In the future, a competitor comes up. And some whiz kid brings out a much better way in algorithm. Wouldn't that lower your competitive ability in the future?

Absolutely. If somebody builds better models than us, it will erode our leadership. I don't think there's a company in the world that I could say that there's no one could ever challenge them, but -- we -- in our view, we have many years head start. We're up full sprint moving into our second category, which I think is probably the hardest thing we've had to do since we initiated is to jump from one category of loans into a second one, which I think kind of paves the path for going to other flavors of credit as well. But yes, no doubt, I'm not here suggesting that no one else could build great models. I am suggesting that we've been doing it for a long time. We don't see a lot of evidence of others building similar models. So we feel comfortable, we have a head start, and we have a great team that's executing really well.

I know there are some follow-up questions here, but we're out of time. We got to give the room up. So if you want to grab Dave on his walking out, that's great. But thank you very much for joining us. Dave, thank you.

Thanks, James.