Upstart Stock – A Play on AI-Powered Consumer Lending
Table of contents
Artificial intelligence (AI) is becoming so pervasive that it now comes in 28 different flavors, each of which has companies claiming to operate in that space. Any entrepreneur knows there’s a bit of “fake it until you make it” going on in all startups, but at some point, you need to show traction. In the world of AI, traction means demonstrating that you can add exponential value. Investors are pouring money into AI because they’re expecting exponential returns.
If you’re a B2C company that offers financial services that come with any risk of loss, you better be using AI to mitigate those losses, or some startup will figure it out for you. For example, auto insurance companies can make more money if they insure drivers who are less likely to get into accidents. Fintech companies like Root Insurance use AI and big data to avoid insuring drivers who are inclined to crash into things. What Root Insurance did for insurance, Upstart wants to do for consumer lending.
About Upstart Stock
Founded in 2012, San Francisco startup Upstart has taken in just over $144 million in funding from a long list of notable investors to develop what started out as peer-to-peer (P2P) lending business. We first came across the company in Spring of 2017 when we wrote about Upstart Loans with Interest Rates Calculated by AI. In that piece, we talked about “the new credit score” which ignores all that FICO rubbish and uses big data instead. Upstart uses more than 1,600 variables to assess the likelihood that someone will default on a loan, and now they’re planning to have an initial public offering (IPO).
The Upstart IPO
The Upstart we see described in the S-1 filing document is a very different one from three years ago when 25% of the $700 million in loans they originated were automated. Fast forward to today and Upstart has now originated over $7 billion worth of loans. In the third quarter of 2020, 70% of their loan applications were approved instantly using a fully automated process (no document upload or phone call required). Over half of the customers that received a loan applied for it on their mobile phone.
Upstart’s AI system is trained on almost 15 billion cells of performance data and predicts the specific likelihood of default or prepayment for each month of a loan term. Being able to describe their loans in such detail means that Upstart can package them and sell them to institutional investors who are eager to buy quality loans that are 75% less likely to default.
Of all the loans originated by Upstart, 78% are sold to institutional investors and 22% are held by the banks that issued the loans. To date, the platform has originated more than 622,000 loans.
Upstart’s Business Model
Upstart doesn’t actually loan anyone money. Their revenue is primarily comprised of fees paid by small to medium-sized banks – referral fees for each loan referred through Upstart.com and originated by a bank partner, platform fees for each loan originated (regardless of its source), and loan servicing fees as consumers repay their loans. In the third quarter of 2020, 96% of their revenues came from bank and loan servicing fees.
As of September 30, 2020, Upstart had 10 bank partners with around 65% of revenues coming from just one – Cross River Bank. While that percentage is decreasing over time, it still represents a risk. A second unnamed bank accounts for another 15% of revenues, meaning 80% of Upstart’s revenues come from two parties right now. Even more risky is the dependency they have on a third party to source leads. More than half of loan originations for Upstart in Q3-2020 were derived via traffic from Credit Karma. That’s up from 38% in 2019, a trend that’s concerning when you read Upstart’s comments about how Credit Karma’s changing business model may mean fewer incoming leads.
Beyond Unsecured Personal Loans
From April 2019 to March 2020, there were $118 billion in unsecured personal loan originations, representing 8% growth over the prior year. In the same period, Upstart originated $3.5 billion in unsecured personal loans, or less than 5% of the total market. While there’s plenty of room for growth, Upstart is also targeting other consumer lending applications – like auto, which is 5X the size of the unsecured loan market. It’s where Upstart plans to move next, and their first auto loan was originated in September 2020.
As you would expect, Upstart has drawn some unwanted attention to themselves around the transparency of their lending algorithms, something they’ve noted as a risk in their regulatory filing. In February 2020, they received a letter from five members of the U.S. Senate asking questions in connection with claims of discriminatory lending made by an advocacy group.
Discrimination in Lending
Individuals who are on the path to success rarely have time to sit around complaining about -isms on Twitter. Still, there’s no shortage of people who will complain about AI’s ability to identify the deadbeats in society who take out loans they can’t afford, use them to buy things they don’t need, then complain when the bill collectors come knocking. When AI “discriminates” against these people by not extending them a line of credit, that’s a good thing.
A common complaint about platforms like Upstart is the notion of transparency. Explain how your algorithms arrived at the conclusions they did, or risk being labeled an evil capitalist who discriminates against <INSERT DISADVANTAGED CLASS OF PEOPLE HERE>. Smart companies ignore this rubbish and the public relations ambulance chasers who try to profit off it. Unfortunately, that’s not an option when the government comes knocking.
As we talked about in our piece on Artificial Intelligence and Transparency, the solution is to stop telling everyone you’re using AI to better mankind. Bury these algorithms deep in the bowels of your back office systems and let them do their magic. If you’re a bank that hasn’t adopted AI for loan-making decisions, you better move quickly and silently to change that.
To Buy or Not to Buy
What Upstart has accomplished is remarkable, but that doesn’t mean it makes for a good investment. The overreliance on Credit Karma and Cross River Bank are a concern, but the bigger picture is what happens when there’s a recession or black swan event. Just look at how quarterly revenues were impacted by The Rona when loan originations plummeted by over 70%.
We’re at a loss for words when it comes to explaining why today’s markets continue to set new highs in the face of unprecedented global uncertainty. It’s probably not a good idea to invest in companies whose fortunes are solely based on the health of the American consumer. When demand dries up for loan originations, especially ones with no collateral, that leaves Upstart in a very difficult spot.
Conclusion
Upstart isn’t a company we’d consider investing in for reasons we’ve given, but they’re an excellent example of how artificial intelligence is being used to improve the quality of life for consumers while making investors a handsome profit. Extending loans to people who fully intend to pay the money back and borrow within their means is good for everyone. Using AI helps us better determine who these people are so we can lend them money and receive a fair return on our capital. With Upstart’s AI algorithms, everyone wins.
If the IPO goes through as planned, Upstart will trade under the symbol UPST.
Share
Strange: not held by ARK. On the other hand: they have LendingClub (LC).
There are any number of reasons why a fund manager might not invest in a stock – liquidity, size, etc. Perhaps they feel that Upstart and LC are an equivalence class and they’re holding the better of the two.
I see a lot of positive articles on Upstart.
Eg: MF : “This Growth Stock Could 10x in 10 Years”.
The only problem is: the price is very high, near 52 week high .. $201
They are already profitable.
Q2 Revenue: $194M, up 1,018% Y|Y
Q2 Net Income: $37.3M, up from ($6.2M) in Q2’20
So I want to have it, but not sure if at the current price ..
The types of articles you refer to are written by pontificating baboons who pull price targets out of their red asses and think they mean something. Focus on the core thesis. You are placing a big bet on unsecured lending to American consumers. If you’re comfortable with that, start adding.
Having not first hand experience of FICO rating, but having followed Fair Isaac business for a while, I would be interested if you guys could expand your thoughts about “FICO rubbish”. Do you think their scoring models are obsolete and could really be disrupt? I have the impression their MOAT is quite solid … cheers, Stefano
The use of the term “FICO rubbish” is meant to be cheeky – from the perspective of Upstart which clearly thinks their method is superior which is why they have a business in the first place.
Thanks a lot for your clarification!