We recently purchased Upstart as part of the momentum portfolio, and we plan to hold it into strength. The company is well known and had a breakout year in 2021. I provide a brief summary of the company’s business model, what we like about the business and key risks.
What does Upstart do?
Upstart is an artificial intelligence (AI) cloud platform that is used to facilitate loan originations. The company uses over 1500 variables in its AI models, which allows its platform to underwrite superior loans with higher approval rates, lower interest rates and lower default rates for consumers compared to legacy lending approaches.
Upstart was founded in 2012 by a Google technologist, Dave Girouard, who worked at Google in 2004, when the company was quickly scaling. Upstart originated its first loan in 2014 and its AI platform more accurately prices risk for unsecured consumer loans, and has now expanded into auto loans. In 2021, 69% of its loan were fully automated, which helps facilitate more loan volumes. Increasing loan volumes strengthen the company’s AI models, improving its competitive moat and giving it a larger lead relative to competitors. Being the first mover in tech is often the most important aspect of the story, and Upstart has a large advantage with its relatively long history of repayment data that will be difficult for competitors to replicate.
Lending is critical to the US economy, and between 2014 through 2021, commercial banks lent out over $16.5 trillion in loans. However, the pricing of these loans has largely relied on legacy methods invented before the emergence of cloud computing and modern data science. A study by Upstart found that four out of five Americans have never defaulted on a loan, yet less than half of them would qualify for the low rates that banks offer. Furthermore, there are enormous amounts of data points available for lending decisions, such as payment data, banking transactions, employment history and educational background. It is a natural progression for cloud computing, AI and machine learning to be applied to lending, given its massive scale and legacy approach.
Revenue model and opportunity
Upstart is a platform for loan originations, its not a bank. Furthermore, the company partners with banks, it does not compete with them like other Fintech companies. The company’s revenues come from fees from originating loans on its platforms that are paid by its banking partners. These fees accounted for 91% of revenues in 2021. Importantly, Upstart’s revenues are low risk and 94% of its sales have no exposure to credit risk.
Revenues are primarily usage-based, and rising loan volumes contribute to rising revenues. As shown below, 2021 was a breakout year for loan volumes, which increased to over $4 billion in Q4 2021. Volumes have primarily been from personal loans that are used to refinance high interest credit card debt for consumers. For example, in Q4, Upstart originated 495,000 loans for $4.1 billion, or $8,300 per loan, and the personal loan market is large, estimated at nearly $100 billion.
The company has fully scaled into personal loans and has now expanded into new markets, such as auto loans. In Q2, the company purchased Prodigy, which expanded its addressable market in auto loans. Management expects auto loans to grow through 2022, which will weigh on its contribution margin in the near term until the company reaches scale, which it expects to do faster than it did with personal loans since it now “has the playbook down”. Beyond auto loans, the company expects to expand into Mortgages in 2023, a massive market at nearly $5 trillion.
What we like about Upstart:
There are a few key aspects of Upstarts model that we find attractive:
– Flywheel effect: Upstart has first-mover advantage, which has allowed it to amass more repayment data and improve its AI models. Its models benefit from a flywheel effect as repayment data leads to improved accuracy of risk pricing, which results in higher approval rates and lower interest rates, which leads to increased volumes and more repayment data. Upstart’s conversion rate, or the number of loans improved per inquiry, increased to 24% in 2021, up from 15% and 13% in 2020 and 2019, respectively. As its AI models improve, its conversion rate should also ramp, increasing volumes and giving the company a significant data advantage.
– 94% of sales have no credit exposure: Upstart is a platform for banks, the company is not a bank, nor does it compete with banks. It offers a marketplace for loans and charges a platform fee and generally does not own the loans. Even if it's 6% of credit exposure defaults, Upstart's high margin earnings can absorb it. The loans it carries are for R&D purposes (discussed below), meaning that its credit exposure should decline going forward.
– Upstart is highly profitable at scale: Loan volumes soared 338% YoY to 1.3 million and total sales increased 264% YoY to $849 million. Contribution margin increased 400 bps YoY to 50%, adjusted EBITDA increased YoY from 13% to 27% and GAAP EPS soared to $1.73. Cash flows from operations increased to $168 million and co-founder-CEO Dave Girouard explained on the Q4 that “We generated more cash in 2021 than we burned in our entire eight-plus years as a private company”. CFO Sanjay Datta added that “the natural profitability of [our] overall model will trend to its equilibrium direction, which we believe is higher than where it is today”
– Upstart’s customers, banks, are outperforming: Upstart's partner banks are doing really well. For example, Customers Bancorp was its first bank partner and Customers' funding form landing page includes a URL link to upstart.com (https://customersbank.upstart.com/funding_formupstart.com/funding_form). Customers Bank was the 2nd largest PPP lender, despite being a relatively small bank, and is a strong lender in personal loans. The company’s partnership with Upstart has likely allowed the company to outperform. Furthermore, high levels of deposits and liquidity levels following the COVID relief programs has created a unique environment for bank lending, and loan demand has been increasing.
– Market opportunity is wide open: Upstart is currently in personal and auto loans. It can expand into the following credit markets: credit cards, mortgages, student loans, small business loans, point-of-sale loans and HELOC. Lending is one of the largest markets ($16T loans since 2014) and is still primarily based on legacy models developed before the cloud era.
– Room for continued improvement: Conversion rates are still relatively low at 24%. As AI models improve from increased repayment data, the conversion rate will rise, leading to more volumes. Also room for continued improvement in fully automated approvals, which were at 69% of total loans.
– Outperformed during COVID: Upstart disclosed in its 10k that during the peak of the COVID pandemic, 5.6% of Upstart’s borrowers had enrolled in a hardship program, less than half of the rate of online industry benchmarks. Furthermore, 95% of these borrowers exited the hardship program and resumed repayments. The COVID-19 pandemic provided valuable data for Upstart, further improving its models and likely also improving the confidence lenders have with its AI powered models. Prior to COVID, Upstart’s models had not been tested during market turmoil, so it was unclear if its AI models were superior. The pandemic may have been an inflection point for the company that proved that its models worked.
Risks:
– Significant customer concentration: Two customers accounted for 83% of total revenues. High exposure to two customers naturally increases the risk of an investment. Furthermore, customer concentration from these two banks increased YoY from 81% of sales in 2020 to 83%. Upstart's largest customer is Cross River Bank, which originated 55% of 2021 loans for the company and also accounted for 56% of fee revenue. Cross River Bank is a partner to many fintech companies and is a conduit between Upstart and institutional investors that will ultimately buy its loans. As a result, the high customer concentration is not as concerning since its actual customer base is much broader. Furthermore, Upstart's reliance on Cross River Bank has been reduced and its second largest customer increased from 18% of 2020 sales to 27% of 2021 sales.
– Cyclical market: Lending is inherently a cyclical market that is impacted by economic strength. If lending stalls, Upstart’s growth will also stall. However, the company is a technology company and while it may be impacted by short term cyclical trends, it is a long-term secular change to the entire industry that should outperform over time.
– Credit exposure to sales has increased from 3% of Q2 sales to 6% of Q4 sales: Upstart is not in the business of collecting interest income, but a portion of its sales come from holding credit risk (loans). Loans carried on its balance sheet increased YoY from $78 million to $252 million in Q4 2021. Management explained on the Q4 call that the large increase in loans was for R&D purposes, as it expands into auto lending. CFO Datta explained that “Most notably, auto lending has been funded since inception entirely from our own balance sheet. This is, as always, a temporary incubation period until we reach the point where the loans can be directed to our bank partners and institutional investors at reasonable scale, which we anticipate will begin to happen next quarter”. However, in the company’s 10K, it disclosed that only $50 million of its $252 million in loans were auto loans; it is unclear why the company holds an additional $200 million in loans on its balance sheet. We will want to see Upstart’s loan balance decline going forward. Luckily, loans >90 days past due are low at just $287,000, so credit impairment risk is low.
Disclaimer: the I/O Fund owns a beneficial ownership in Upstart. The I/O Fund did not receive compensation for authoring this article from any of the discussed companies,