Upstart Holdings (UPST -10.21%) is a lending platform, powered by artificial intelligence. Its shares are down more than 70% year to date. It’s down, but not out.
In this podcast, Upstart CEO Dave Girouard joined Motley Fool CEO Tom Gardner to discuss:
- How Upstart is using its balance sheet now.
- Growth opportunities in auto lending.
- One stock idea (that’s not his own company).
To catch full episodes of all The Motley Fool’s free podcasts, check out our podcast center. To get started investing, check out our quick-start guide to investing in stocks. A full transcript follows the video.
This video was recorded on June 19, 2022.
Dave Girouard: Also, even whatever is on our balance sheet, it’s a fraction of what a lot of other fintechs have, we’re not a balance sheet company, do not intend to become one. Maybe we just need to be a little more disciplined and have technology that’s a little better at price discovery. I think that’s what we’re getting toward.
Chris Hill: I’m Chris Hill and that’s Dave Girouard, the CEO of Upstart Holdings, a lending platform that uses AI to determine creditworthiness instead of the traditional credit score. The company ran into hot water when some investors were less than pleased to find loans on its balance sheet; year to date shares of Upstart are down more than 70%. Girouard joined Motley Fool CEO Tom Gardner to break down how Upstart is using its balance sheet, growth opportunities moving forward, and one stock idea outside of his own company.
Tom Gardner: Now I want to talk specifically about Upstart, the business, and then really get to today’s environment in the latter part of the conversation. But right now, as you look at your company, what do you see at the top few competitive advantages?
Dave Girouard: We’re going after a very hard problem that I think very few others are even concerned about or attempting to deal with and that is access to credit. The simplest way I can describe it is, in our view, 80-90% of Americans are fundamentally creditworthy, given the right product at a reasonable price, they will pay back that loan, whatever form that is. But really, only about half of Americans are recognized as such. There’s just an enormous difference between the reality of the risk in the world in how these very archaic systems we have to measure it work. People might just say, yes, that’s the world and that’s how it is. Some people got good credit scores and some don’t. But there is a better reality out there. I am a technology optimist. I’ve grown up in the technology world. I’ve seen the way it can transform industries and worlds in just unimaginable ways.
Yet somehow are we supposed to believe that this notion of having more accurate credit and making credit more readily available to more people at better prices is impossible to solve? That’s the heart of what we’re focused on in that building the application of artificial intelligence to credit, to us is an obvious join. AI is clearly a technology that has amazing potential in so many dimensions. Credit is a risk-based problem that is vast in scope. If you can be the company that really leads the charge in terms of applying AI to this enormous giant industry, the potential there is awesome. It does come down to execution. There are a lot of pieces and parts of the problems to solve. But the addressable market or the chance you have to do it is so vast. Honestly, we’re not like Uber and Lyft elbowing each other or we don’t have others that are really in our face, if you want to say, trying to do the same thing. Most are just happy to build a digital bank or some other type of payment company. All great businesses, other businesses, but there are really very few looking to innovate on this very specific problem of how you make credit work better and more efficiently and through the application of very sophisticated models. That’s a place we feel, particularly in the U.S. market, really, in a class by ourselves.
Tom Gardner: Perhaps you just answered this, but what would you say to somebody who said, “Upstart has created something that would be a useful application inside of a larger bank, but I don’t see it as a stand-alone company”?
Dave Girouard: Yeah. I think that’s probably an inside-out way to look at it. Any particular bank solves a problem for a particular set of customers and a certain set of products. The technology that we’re developing here is utility far beyond any single bank and what they would want to do. There’s a good reason to be a bank. Some fintechs are deciding to become banks. You have your own balance sheet, you have deposits and you lend against them and it’s a known thing.
If you’re a new age 2022 digital bank, you have some advantages. That’s all good, but in our view the most impactful thing you could do is to insert this technology into as the highest fraction of credit originations of all flavors going on in the United States and eventually the world. You can never do that as a bank. You could be a great bank, but the impact in the scale that you can build at, if you are a tool that every bank, every credit union, every lender can use, I believe there’s a much more scalable, much more impactful model. One that is, as a guy that came from [Alphabet‘s] Google, a bunch of folks came from Google and companies like that, that’s the appealing opportunity we’re excited about.
Tom Gardner: What did you just learn in deciding not to take loans on the balance sheet as a market clearing mechanism. I guess, in a way, I was wondering or thinking that maybe it was almost like putting your personal lending business back in R&D because the environment changed so substantially that the model needed to catch up to the new reality. But you’ll see and I’m sure you have seen or your investor relations team has seen littered around the internet, comments saying, well, they said they were a tech company, but they’re a bank, they’re a subprime lender now. They’re taking a lending risk and it’s just beginning because if this environment worsens then do that even more and obviously now you’ve changed course on that. You talk about the importance or significance of Upstart not becoming a bank. There are people who believe that you just took a step in that direction. Correct that thinking or explain your rationale today.
Dave Girouard: Yeah. We’ve always grown up as a balance sheet company not intending to build loans on our balance sheet to generate net interest income, which is again, a perfectly good business, but not our business. As we’ve said before, our business is in effect a marketplace. If you wanted to say that technology is one thing, the business model’s a different thing. The business model is largely a marketplace with consumers on one side, banks and lenders and investors on the other side. What we have said and continue to say is we will take things on our balance sheet to test and try out new things. It’s really a form of R&D. But ultimately beyond that, we want to be a market-maker. Now, like in any market, you can have surpluses on the supplier-demand side, the pendulum swings back and forth in any kind of marketplace business.
The truth is on the funding side, it’s just more brittle and it’s not as getting price discovery, making supply meet demand, which is, of course, the objective of any marketplace. It’s not as fluid as we would like. Some of the things we can fix and some of them may just be endemic to the nature of banks and lending and capital markets investors, they react emotionally sometimes more than just to numbers. But in any case, the bottom line is, we move, I would say 85-ish percent of the time we’ve been in this business, we have been borrower constrained, meaning unlimited sources of lenders and funding and always just borrower constrained in terms of where we can economically bring borrowers onto the platform. For the other 15% we found a place where there’s overwhelming consumer demand, there’s not enough lending capacity out there, which is where we’ve been.
In March, we made that switch happen pretty quickly for a whole bunch of reasons that are largely about unfortunately war and inflation and things of that nature. But in any case, we were caught on the loss side. We weren’t as good and aren’t yet as good at getting price discovery happening, meaning prices move up until supply meets demand. That’s maybe the economics 101 thing that needs to happen. That’s our intention going forward. Also, even whatever is on our balance sheet, it’s a fraction of what a lot of other fintechs are. We’re not a balance sheet company, and do not intend to become one. Maybe we just need to be a little more disciplined and have technology that’s a little better at price discovery. I think that’s what we’re getting toward.
Tom Gardner: Just to understand the thinking that went behind that, was that to fill in a gap in the marketplace for defensive reasons or for revenue-generating reasons? What ticked you toward making that decision?
Dave Girouard: It’s just continuity. The pipes are running, the borrowers are applying, they are being matched to lenders, and the lenders are selling some of the loans, keeping some of the loans. So, I mean, it’s just sometimes like in March 2020 to go back a couple of years, there was just an insane upset of the applecart in the course of a few weeks. This wasn’t quite like that, but you just have that when the economy changes really quickly. We have to make decisions really fast on such things, and generally speaking, like I said, that’s not our goal. Interest income is not of interest to us.
We really aim to be the marketplace and the partner to these banks and credit unions. I think we’re going to get closer and closer there. The important thing I’d say is we have mechanisms to make sure supply meets demand. For every loan that is approved on our platform, it is known what bank is originating it, and if that bank is not going to hold themselves then what investor will have it after the fact. So there’s never a case where there’s a loan sitting around and, “We’re left with them.” But it was a decision really to keep the pipeline moving and not to defer it, and we have a lot of tools to make sure supply meets demand. We’re going to get better at that to make sure we’re good with that in the future.
Tom Gardner: You talked about growing a company over long periods of time as a risk mitigation exercise. Which do you feel is the bigger risk to Upstart if you can compare these two? Would you say the bigger risk is that your data advantage of 10 years is not as great as you had hoped because other people came along, leapfrogged, and there were different sources of data? It wasn’t as big a lead as you thought you had in the last 10 years. That’s one or two, that the 10 years of data you have is in a low rate interest rate environment and the models, your fear about the adaptability of the model when conditions worsen when credit markets stall. If you compare those two, which one do you think is a greater long-term risk?
Dave Girouard: Honestly, I don’t fear either of them because I don’t think there’s any evidence of either. We don’t see others building models similar to ours. The best thing we can do is try to observe how other models work, and they all, in terms of consumer lending, all tend to be so highly correlated to a credit score that it’s really hard to see anything beyond that going on. Maybe around the edges, but we just don’t see it. It’s hard for me to worry that suddenly our advantage out there is lessening. I don’t see that. On the second thing, I mean, conceptually you could worry that your model, the environment is going to change such that your model suddenly becomes useless if you want. For us, it’s almost implausible to imagine that because again, it’s how it’s doing relative to a traditional credit score, and that’s not a tough fight for us just to be frank, like the amount of risk separation, if you just look. We’ve put actually a slide about this on our investor deck.
If you just split all of our loans by credit score and then you split all over the same loans by the Upstart, essentially risk tier, what you see is a dramatic separation and the risk tiers. A very smooth from tier 1 up to tier 8 like a very smooth increase in loss rates as you would like across these risk tiers. The tiers are working incredibly well. Whereas FICO it’s only lightly correlated. It’s useful a little bit, but it’s actually not that well correlated. But anyway, I don’t want to sound like we don’t have things to worry about, Tom. Every business does and we do. I feel like if I can just nominate a No. 3 three, we have to execute. There are a lot of things that can go wrong in any particular business, and certainly in ours. For us it’s execution to grab the opportunity to prove this isn’t about unsecured personal lending, it’s also about auto lending, it’s also about small-business lending and mortgage lending. So to prove them more to categories to win over more lenders to the platform. Those are the things I worry about is really, how do we take those next steps to really prove this is going to be the business that we believe it’s going to be in a few years.
Tom Gardner: Let’s go to the other categories. Let’s move to auto-lending now and enlighten us, teach us all the differences between personal lending and auto lending. The size of the market, the competitive dynamics in those markets, the potential margins, and the amount of market share that you think is available to Upstart in the two different categories. Those two to start.
Dave Girouard: Personal lending, depending on how you measure it, it’s maybe a $100, $150 billion a year in originations. We believe our platform is a potential market share leader in the U.S. in that category and has become so over the last few years. But it’s not a mainstream credit product, meaning most banks don’t really offer personal loans at scale. It’s historically an esoteric product just because it wasn’t very economic for banks to like make a $10,000 loan. They just weren’t going to make enough interest on that to make it worth the effort of dealing. Fintechs have really almost created that category in the last decade. We’ve really built a very strong position there and continue to build on that. Auto is very different of course, it’s a very well-established category. It is probably scale-wise, maybe seven, eight times larger. Maybe it’s a $700 billion a year in origin, maybe $800 billion a year in origination, so much larger, much more mainstream to the financial services world, to the banking world. This is a secured loan, so it’s a fundamentally different product, whereas unsecured personal is really like you’re betting on the person, you are underwriting the person. In an auto loan, much like a mortgage, there’s a person, but there’s also an asset behind it, and that means it’s a collateralized loan. There’s something you can recover.
Generally speaking, it makes the loan notionally less risky because you can recover a higher fraction if the person chooses not to pay. Getting that right and how that works. Also, the payment waterfall is different. Generally speaking, somebody is more likely to not pay a personal loan where the quid pro quo is they mind their credit score might get hit or a car loan their car might get repossessed, and so generally speaking, it’s always believed that a car loan is higher up the payment waterfall for the consumer. That’s how they differ.
From our point of view, there is still very sophisticated modeling in both. There are a lot of processes involved in a car loan that are not involved in a personal loan. If it’s a refinance, which is the first product we got in, you have to deal with paying off the prior lender, establishing the new creditor on the lien, on the title for the car, so there’s some logistical stuff that can make it what I would call a historically 0 billion-dollar market. Meaning, if it took 10 minutes, who wouldn’t want to refinance the car loan to save a couple $100 a month? But if I had to go through trudging to the DMV and God-knows-what and notarization and all that, maybe I just won’t bother. I think that’s where the industry has been to date. So we’ve been building a process that feels more like unsecured personal products.
It’s all automated, it can be done really quickly. I think even the bigger opportunity for us in auto is at the car dealership itself when people are buying cars. Historically, one of the, let’s just say, worst experiences ever invented in the United States of America is what most people experience when they go to buy a car. It’s just a circumstantial thing that’s built over time, but we bought a company a year or so ago now, Prodigy, that is really the software going into car dealerships to help them create a more pleasing process for all, a more efficient car buying process. We’re just now testing Upstart loans in that process. That’s an enormous opportunity because that’s where the bulk of auto-lending happens, the vast majority. It’s really efficient, both in terms of process and in terms of pricing. It pulls on both of the ropes that make our business go, and we’re seeing extremely promising early results. Our view generally is if we were betting on this, in a few years, you’d see auto surpass our personal loan business just by the potential of the inefficiency, and what we think is a very good position. We have, we feel pretty confidently, the fastest-growing auto retail software that’s in the industry. There’s a whole bunch of providers of trying to make software that helps car dealers sell more efficiently. But ours is clearly growing faster than others.
Tom Gardner: What do you expect the margins still look like in auto-lending versus personal lending?
Dave Girouard: Margins to us, we think won’t be very different, very similar. Generally speaking, more of the revenue will likely happen over the term of the loan as opposed to upfront, so you can view that either as a good or bad thing. But we think the margins and the take and all that won’t be all that different. I think the level of inefficiency and opportunity, are pretty similar. But I do think the nature of that product isn’t a large upfront fee or anything like that, so it will tend to be a bit more recognized over the term of the loan, which for the point of just stability if you will, is not necessarily a bad thing.
Tom Gardner: We’re going to take a step back for everyone to just get their footing about AI and to have you explain taking the auto loan on your balance sheet as an R&D maneuver to train the AI, so as to make sure the system works so that now lenders can come into the platform and feel confident that the data you’re presenting is valid. So could you just explain that process and how somebody just coming into it for the first time might say, what? Wait, all these loans are on their balance sheet. This isn’t a risk-less organization, they’re having to shoulder all of these loans and it’s in the hundreds of millions of dollars. Walk us through the process of how that works.
Dave Girouard: So just to give you an example. So we have a small business loan product coming out later this year. We’ve talked about it a bit, it’s on the near-term horizon. For the rest of this year, we’ll probably hope to really get this thing tested and out there and try it. Maybe a couple of tens of millions of dollars of loans, which in the grand scheme of lending is not a lot. Now we can’t go to one of our bank partners and say, hey, we’ve got this thing, it’s ready to go, you want to get the first small business loan in Upstart? Because that just doesn’t make sense for them. They have a lot of responsibility in their vetting on something. We just never allow for that. So when we’re bringing a new product to market or maybe something very different in an existing product, we want to have the capacity to test it ourselves and get through the first version, the second version, maybe the third version of the model. Usually, the curves are such that you can iterate quite quickly if you have enough volume. That’s what we’ve been doing in auto, is really funding most of it and testing it.
Now for the refi product which has been in the market for a while, we’re now transitioning that, where banks and credit unions are becoming the lenders for that, and we’re getting it off our balance sheet. That’s how it’s supposed to work. We do it for, it depends on the product, it might be for six months or nine months or something like that, maybe a year. Then at that point, lenders have enough confidence in the product that they can step in and take it and be happy with it and it can pass all their tests if you will. Then small business now. We’re going to start that process for that product soon enough, and we’ll go through the same thing, six, nine months, who knows exactly? That capital on our balance sheet, it’s an incredibly valuable use of it because it is R&D. I don’t know how you would build an AI model where the bank assumes all the risks of the learning of getting this thing right in figuring it out. That’s not a reasonable position for a bank to take and it’s not a reasonable ask for us to make so we don’t do that. We said look, we’re going to build the first version of this ourselves going to test the pipes we’re going to refine the model, and when you’re comfortable, you come on board and then we’ll be ready for you.
Tom Gardner: I’m sorry to go back to this, I promise this will be the last time I ask about taking the personal lending business back on the balance sheet. But is it fair to say that that was happening because essentially you needed to put it back in R&D to show the banks that this model does work in the new regime?
Dave Girouard: I think that would be a decorative description of it. In reality, I think it was just a mismatch of supply and demand that happened in a very short period of time. We just opted not to turn off some of the pipes as quickly as we could have, which would’ve been a different choice. Really it was just sometimes you don’t know if something is momentary and it’s going to clear itself up in a few days or whether there’s something deeper going on in the economy or whatever. I would like to say, I can’t say that was a form of R&D, and we would pretty transparent that that’s what happened is suddenly the markets did turn and our price discovery process isn’t fast enough to get prices where supply meets demand, and when we had that disequilibrium, if you will, we took some of them onto our balance sheet, and we do not intend to do that, it’s not our business, and we’re going to get better at the tools to have price discovery happen faster.
One of the very encouraging things we’re seeing is we have a lot of pricing power. We haven’t moved prices up a lot, I said this earlier, because core interest rates moved up, Fed rates moved up, probably the two-year treasury is really the mark that matters the most to us, and the two-year treasuries up a couple hundred basis points since the fall, as well as the risks and the environment that also pushes rates up. Yet still, consumer demand is super strong and that just shows we have, I think, real pricing power, which is good. But it means we’re not as good as we want to be at getting price discovery happening.
Tom Gardner: We live interesting times. Certainly, a comparable would be to go back around 20 years and see what happened to the valuation of a lot of technology and growth companies. In hindsight, say their prices got well ahead of their value, but then the best among them delivered some of the greatest returns in American market history after that because the actual revolution was real, it was tangible even though there were a bunch of joke companies should never have even existed, let alone come public.
There were actually some, obviously an amazing companies and every company we’d like to compare itself to Amazon and the public markets, of course. But in that, I think 2001 shareholder letter, Jeff Bezos I think the word “ouch” was right there in the beginning. Our stock is down 90% and Bezos in talking about at some interviews I’ve seen it after the fact said it was funny because internally a lot of things we’re going exactly the direction that we wanted them to go. But maybe a collection of valuation, reset, big new changes in the environment and trouble communicating what we were achieving and going for all all hit us at once. But when I was looking at the internal numbers, I was actually very pleased by what was happening, but our stock was down 90 percent at the time. Without putting you in a position where you have to compare yourself to Amazon, given that, compare and contrast the feeling that you have right now inside of Upstart, what you’re seeing develop at the company versus the absolute invalidation of any prior valuation to a $150 a share all in the six-month periods slamming you. The external validation is getting knocked down and the internal experience, how different is that for you?
Dave Girouard: It’s not as stark it might feel to the outsiders as you might think we’re just like demotivated or just crushed by this. I don’t think there’s a lot of that. I wouldn’t say nobody looks at the stock price. That would be silly. But honestly, we’re pretty focused on the mission on what we’re accomplishing internally. I think a good lesson for one of your members if they really want to understand us, forget all the noise, forget the stock price at any point in time. Go to the beginning, at least as our public journey, read the S1, read how we describe what we do, why we do it, how we do it, because we put a lot of energy in our S1, going way back then to describe how this AI works, it’s not just noise there, there’s some deep science there and we actually got super disclosive of it in the S1.
Then go read our first earnings report, our second earnings report, our third, and just take all the market noise out of that and judge for yourself is this real? Are these people legitimate, do they do what said they’re going to do. If you do that at whatever conclusion you come to you can come to. But I think in some sense you have to ignore the fact that the stock, which started with $20 by the way, when we went public, ran up to close to $400, came all the way back to wherever it’s sitting today, $45 bucks. [laughs] But again, that’s the market, that’s the noise. Read the details of what we’ve done and who we are, and I think you have a better chance to get to the truth than just reading speculative ideas about what we’re good at, and not good at we said this, and we’re not perfect, we’ll make mistakes.
Any decent company trying to be transformative and trying to do really hard things is going to make mistakes, and we’re in that list, but we’re also in a very strong place, a very strong position to launch these new products from a really well capitalizing. We’re company I just by the way, Tom, as a private company, we raised a total of a $160 million, frankly, a fraction of most fintechs. When we went public, I think we had in the range of $90 million of that still on our balance sheet. We’ve just been that company since day one, that’s never changed. I think if you want to be a long-term investor, you got to get to the heart of a company, who they are, who the people are, how they do what they do, and then place your bets.
Tom Gardner: This is probably my least favorite question to ask. One of those rash statements that floats around it maybe not floats around, that spirals around, and I want to give you an opportunity to explain how the process works on executive selling of shares. Because these stories, your stock that has about a 30% short interest, and therefore, I don’t want to be conspiratorial in my thinking, but therefore there are group of people because shorting is a short-term transaction that have a short-term incentive to swirl some rumor out there in the marketplace. Could you talk about your ownership stake, shares that you’ve sold in the last year, how that works and what it says about your commitment or a lack of commitment to Upstart.
Dave Girouard: Let me give proper context to it. We’re eight years as a private company and now 8.5, as a private and 1.5 as a public. The 8.5 years as a private, nobody, no insider sold a single share. In fact, there’s at least a couple of junctures where nobody wanted to fund our business, honestly, and a couple of times or I put what amounted to pretty significant parts of my personal worth, my family’s worth into the business to get it to the next step and nobody sold anything, not me, not our board, not any of the executives. The only selling that’s gone on since we became public from executives has been through 10b5-1 plans, structured selling plans where there’s setup in advance and you have no choice. My plan was set up over a year-ago, May 2021, to sell what amounts to a single-digit fraction of what I own based on price triggers, etc, those things always are, and that’s it. I have no ability to change that. Can you legally stop them or not? I don’t know, but they were set based on what the world looked like and what Upstart looked like in May 2021, and that’s it. That’s the long and the short of it. I own the vast majority of shares I’ve ever had in Upstart, and I expect to have them for a very long time.
Tom Gardner: Last question, which is probably a one you wouldn’t expect me to ask to close, but if you could, in a very generalized way, provide investment advice to investors in high-growth technology enterprising companies, given what you’ve seen previously at Google, watching what happened in 2000. 2001, 2002, 2008, 2010, and a different sudden drops, but this is obviously substantial one when you have the Nasdaq falling more than 25 percent, that’s maybe a once out of every 10-year outcome or experienced for the Nasdaq. What advice do you have for us as investors in companies like yours, not specifically Upstart? But just if you’re investing in companies that are spending on R&D and trying to explore the future, and their stocks have gotten rocked 30%, 50%, 70% or more, what advice would you have for anyone who’s thinking about their portfolio now and seeing a lot of red?
Dave Girouard: Obviously you don’t want to act in fear, I’m one who has not historically done a lot of singular stock picking. I do it occasionally, but I usually will not do a lot of that myself. But occasionally I just have conviction and I have conviction through experience in seeing a product, and I will just give you an example. I put a big chunk of money recently, the first time I bought a singular stock in a long into Zoom. I was I know that business, we’re trying to build products like that at Google. I know how hard it is. That company executed incredibly well when suddenly their business just went through the roof in early 2020, and I had just so much respect for what they’ve done, and I know how hard the problem is to solve. How many times has like doing video like this has been just a nightmare in the past despite the fact that Microsoft’s coming after them, Google’s coming after them whoever else.
To me it is you can do index investing or whatever you want, but you want to have some conviction somewhere. I don’t know if I got Zoom at its lowest or whatever and timing the market is just not a useful exercise, I think. But find an area where you have conviction. You’ve seen what a team can do, you have enough personal experience to know it, such as something somebody mentioned to you and that’s how I think about it. Honestly, I invested in Apple [laughs] in 2001 because I thought the iPod was a pretty damn awesome concept. I thought, wow, Steve Jobs can create a No. 1 position. I had left Apple a few years before that, and I fairly disgusted with the company when I left it. I said, if you can do that with an iPod, what else is he going to do over the time, and that one worked out pretty well.
Chris Hill: As always, people on the program may have interest in the stocks they talk about, and The Motley Fool may have formal recommendations for or against, so don’t buy or sell stocks based solely on what you hear. I’m Chris Hill. Thanks for listening. The market is closed on Monday for the Juneteenth holiday, so we will see you on Tuesday.