Upstart (UPST): Building The Lending Platform Of the AI Era

Many financial services are, at their core, information businesses. For example, an insurance company needs to have correct information about the future rate of accidents and their cost to properly identify the price at which an insurance policy will still be profitable.

Similarly, lending requires properly assessing the risk of default from a given customer, and setting the interest charged at the appropriate level.

Historically, this was first done with sole human judgement and primitive mathematical formulas, with social class, personal connections, and personal behaviour impacting an individual’s access to financial services.

More professional and objective methods progressively emerged to properly assess risk with less bias. For example, credit scores can become, in America, the dominant form of evaluation for assessing an applicant’s borrowing capacity.

A modern credit score is based on five distinct data categories:

• Payment history (35%)
• Amount owed (30%)
• Length of credit history (15%)
• New credit and inquiries (10%)
• Credit mix (10%)

But of course, it has its limitations. For example, it does not include any information about an individual’s behavior outside of things related to past and present debt. It also does not consider information like place of living, income, sector of employment, assets owned, etc.

So more advanced systems could be built, especially as applicable data are abundant and more readily available in the information age.

This is around this idea that Upstart Holdings (UPST ) was built. Using an early form of machine learning and AI, the company built a more accurate lending platform able to deliver loans to more people while maintaining a lower default rate.

This head start has put the company in an ideal position to benefit from the recent progress in AI technology and build an ever-expanding and more advanced AI-driven lending and financial services platform.

“As an AI-centric company, our business grows when our models get better at separating and calibrating risk. These model improvements help us approve more customers for more products, which improves our business, and puts more distance between Upstart AI and the rest of the industry.”

Upstart Overview

Upstart’s History

Upstart launched in 2012 as a lending platform based in California by ex-Google employees and a Thiel Fellow. Its first product was an Income Share Agreement (ISA), which enabled individuals to raise money by contracting to share a percent of their future income.

2 years later, the company pivoted toward the unsecured personal loan market instead, starting with traditional 3-year loans.

The technical background of its founders led the company to immediately create an early form of AI instead of a traditional credit score. It used data such as educational history, field of study, GPA, and employment history.

From 2014-2019, the company raised significant funding from major institutional investors like  Google Ventures, Mark Cuban, Eric Schmidt,  Marc Benioff, Rakuten, etc.

It went public in 2020 with a price of approximately $20 per share.

The stock immediately shot up, propelled by low pandemic-era interest rates and high demand for credit, ultimately hitting an all-time high of $390.00 / share in October 2021.

This quick stock rise was, however, followed by a brutal downfall, as the Fed aggressively raised interest rates through 2022 and 2023.

This led to a painful transition toward a high-interest-rate environment where institutional loan buyers dried up, defaults increased, and the stock hit a low point of $11.93 on May 3, 2023.

Since 2025, the company has reverted to profitability and restored a strong growth trend, with revenues up 64% year-over-year in 2025.

Upstart By the Numbers

In 2025, Upstart recorded a banner year with large double-digit growth numbers year-to-year on most metrics, from loan origination (+86%) to total loan number (+115%) and total revenues (+64%).

This trend is still going on strongly in Q2 2026, with $0.2B in origination (+50% YoY) and total revenues of $365M (+42% YoY).

Source: Upstart

The company employs 1,405 full-time employees, of which around 430 are in engineering, to train and refine its AI underwriting algorithms. It practices a digital-first strategy, allowing staff to work from anywhere in the U.S., backed by 4 primary physical office hubs (California, Ohio, New York, Texas).

Upstart holds 4 active U.S. patents protecting its proprietary financial machine learning risk models.

Upstart is mostly a loan rating & origination company, and holds only 5.9% of total outstanding loans ($1.06B) directly on Upstart’s balance sheet. The remaining ~94% are funded by institutional investors or partner credit unions.

Most of the company’s business is centered around unsecured personal loans (62% of total loans), but the new segments like auto loans and mortgages are quickly rising.

Source: Upstart

Upstart’s Business Model

Beyond Credit Scores

The key initial insight for Upstart was to figure out that the FICO score is just a crude approximation, and that a customized risk grade taking into account more personal data, but also macroeconomic indicators, could perform better.

And it quickly did just that, with the ability to distinguish within a homogeneous FICO score group which customers are actually low or high risk.

“Our AI underwriting model is 2.74X as good as a traditional credit model”

For example, a 700 FICO Score individual will be reclassified into Upstart’s risk grade to correctly predict the real annualized default rate of his group, which can range from 1.1% to 13%.

So it is not to say that Upstart does not use the FICO score, as it provides valuable information in itself. But the intersection of FICO and its own risk grade creates a much more granular and precise evaluation of lending risks.

Source: Upstart

In addition to lower risks, using an algorithm instead of a loan officer means the process is automated. This is both less expensive to do and much quicker and more pleasant to apply for a loan, with the whole process online.

Source: Upstart

Upgraded AI Model

To determine individual loan risks, Upstart now uses its Model 19, an advanced AI-driven underwriting model. This is an 88% more complex model than its previous versions. It analyzes more than 2,500 granular variables to determine exactly which consumers receive those higher interest rates.

Source: Upstart

It is also a major upgrade in its base design: previous systems looked at the final outcome of a loan—whether it was fully paid off or ended in a charge-off (default). Model 19 instead tracks and models the probabilities of a loan transitioning between intermediate statuses (e.g., current, 30 days delinquent, 60 days delinquent, recovered, or paid off) month by month.

This helps Model 19 identify, for example, when a loan gets one month past due but successfully recovers to “current” status. Cumulated over millions of loans, it creates a much richer dataset which drastically improves risk prediction accuracy.

More Macroeconomics Data

The credit score risk is combined with the Upstart Macro Index (UMI), which estimates the impact of the macroeconomy on credit losses for Upstart-powered loans.

A normal macroeconomic environment is represented by a UMI of 1.0; a higher score marks a higher risk of default than during normal economic cycles due to inflation and financial headwinds (UMI is standing at 1.5 on September 3, 2026).

Source: Upstart

When the monthly UMI calculation moves, the Upstart system automatically recalculates the loss assumptions for all new originations and does not wait for default to first rise. For the consumer, this translates to higher Annual Percentage Rates (APRs) or outright non-approvals to protect institutional investor capital.

This is not necessarily a totally new idea, as some banks already had a similar idea in place internally, but it represents a step above relying on only individual criteria, or just reducing loan volume during economic headwinds not yet impacting default rates.

Upstart Future

Upstart’s head start did not just translate into more operational experience in leveraging AI to better predict loan outcomes. It also let the company collect a treasure trove of data, with the volume of training data points growing steadily since 2018, now reaching the tens of millions.

Source: Upstart

Upstart’s AI advantage has translated into keeping risks low while also providing its customers with the lowest interest rate in around 45-50% of the time, way ahead of its nearest peer.

This position in AI-controlled lending gives Upstart a perfect position for a decade-long runaway of growth ahead.

The company serves an absolutely massive market, with $25T in credit originated each year globally. Of this sum, up to $1T in revenue is available to the platforms that originate and service that credit.

The company is also growing quickly in new segments, with, for example, auto loan origination growing 5x since Q4 2024 and home loan origination growing 6x.

So even if apparently large in absolute numbers, the current segment of the market captured by Upstart is actually still very small compared to the whole.

The recent growth numbers and having been built as an AI-first company are strong positives for the future of Upstart.

The recent success in home and auto loans indicates well that the methodology of Upstart could be expanded to every loan market. This, for example, would cover student loans, subprime loans, small businesses, corporate loans, bonds, trade finance markets, asset-based lending  & equipment finance markets, etc.

Source: Upstart

An AI-first approach could also be brought to innovative financial services, like peer-to-peer (P2P) lending and the DeFi (Decentralized Finance) crypto lending market.

Another strength of Upstart is the easy application process and better approval rate/interest rate offered, which create repeat business. This lowers marketing costs and increases revenues organically.

“Borrowers are coming back for another loan faster than ever. The 2024 and 2025 cohorts are borrowing again at roughly three times the pace of any earlier vintage.”

Source: Upstart

Investors Takeaways

Despite its recent return to profitability and explosive growth, Upstart’s stock price has not really rewarded this success, with prices at the same level as in 2022, and much below its 2021 peak during the pandemic.

In part, this is because a lot of the attention of the market is centered around AI models, AI hardware, and less around specialized applications of AI.

But this is because there are also a few risks to Upstart’s business model.

The first risk is that its business model can deteriorate, as economic conditions move beyond training data and its AI could miss new trends.

This is also generally a risk for the loan industry as a whole, as economic shocks often create rising default rates above what the theoretical model predicted, as exemplified by the 2008 Financial Crisis. With a looming energy crisis, war in the Middle East, and the Ukraine-Russia war, this is not a risk to dismiss too quickly.

Another issue is increasing regulatory pressure for explainability, fairness, and demographic performance reporting, as even if inadvertently, AI-controlled lending could bring back discriminatory practices previously banned from the financial industry. For example, some of the data used could correlate too closely with an individual’s characteristics like gender, race, sexual identity, religion, etc., which are protected from being used as the basis for a loan decision.

Lastly, the company still remains dependent on institutional funding remaining available across credit cycles, with signs of stress in the financial system building in the context of high debt levels and a tense geopolitical situation.

However, Upstart is ideally placed to become a key AI lending marketplace and underwriting infrastructure provider. So investors might want to get early exposure to a company that has been radically changing how loan underwriting is performed, and quickly becoming a giant in loan origination, credit risk measurement, and automated loan assessment.

This means that short of a serious external shock (and this risk stands for most assets), Upstart’s investing profile could provide potential shareholders with a good mix of value and growth, with a price discounting the growth due to past troubles, and a remarkable triple-digit percentage growth in revenues.

Latest Upstart Holdings (UPST) Stock News and Developments