Third quarter earnings call, November 3, 2026
Upstart reported a quarter in which the origination engine outperformed and the balance sheet dragged. Contribution profit reached a record $218 million, up 48% year-over-year, yet management lowered full-year total revenue guidance to about $1.35 billion from $1.4 billion and adjusted EBITDA to about $260 million from $294 million. The company still expects fee revenue and contribution profit to finish modestly above the original outlook, so the guidance revision is mixed rather than uniformly negative. It is, however, a cut to the two headline lines most investors track, and the cause is a macro sensitivity that management is only beginning to reduce.
The Balance Sheet Is the Problem, and Management Says So
CEO Paul Gu framed the quarter with the formula "growth of our business is equal to alpha plus beta times macro." Alpha is the growth the company earns from its models and marketing. Beta is the sensitivity of its results to the macro environment, which Gu acknowledged is high on the balance sheet. The Upstart Macro Index, the company's gauge of macro-driven default risk on personal loans, averaged 1.50 in the quarter versus 1.41 in Q2 and stood at 1.53 at the latest reading, up from 1.39 when guidance was set in February. Over the same period the 2-year Treasury yield rose more than 115 basis points, with roughly half of that move in September alone.
Management's rule of thumb is that each 1-point rise in UMI, or 10 basis point rise in funding costs, cuts daily originations by 1% to 2%. The more important point is that this heuristic covers only the fee business. Gu said the impact on loans and co-investments held on the balance sheet is larger: "Because these positions were accumulated over many quarters, they are relatively large. As a result, an increase in UMI causes a significant reduction in fair values and net interest income. The impact to total revenue and profit from this is actually larger than the impact from reducing new originations."
The numbers show it. Net interest income and fair value adjustments totaled about $6 million, down 65% year-over-year and down roughly $10 million sequentially. About half of the sequential decline came from fair value write-downs and charge-offs tied to the higher UMI, and the other half from a Q2 debt sale that did not repeat. CFO Andrea Blankmeyer noted that the fair value line on the company's beneficial interest assets swung from a positive $13 million in Q1 to negative $7 million to $8 million in each of Q2 and Q3. The Q4 guide implies a further sizable negative mark, mostly on beneficial interest assets, and the company allowed for "some degree of variability." In response to an analyst who pushed on why a UMI move of only a couple of hundredths implies such a large Q4 swing, management explained that realized delinquencies and charge-offs are building from the elevated UMI seen across most of 2026, so the marks reflect accumulated pressure rather than only the incremental move.
The guidance assumes UMI does not exceed 1.55 and that Treasury yields do not change materially. Management declined to quantify what a move to 1.60 would do, saying only that the flow-through is "highly correlated" but not direct. That leaves a real unquantified downside to the balance sheet lines if consumer credit keeps deteriorating. Gu cited inflation reaccelerating to 3.7%, prices rising at twice the pace of wages, and the share of Americans at least 30 days late on any credit obligation rising above 19% in September, the highest in years.
Origination Momentum Slows in October
Q3 originations were $4.1 billion, up 44%, with unsecured at $3.5 billion (core personal loans about $2.45 billion) and secured at $587 million, up 194%. Fee revenue was a record $364 million, up 41%, and the take rate rose about 60 basis points sequentially to 8.8%. Management attributed the unsecured gain mainly to a modest mix shift toward core loans, and the secured gain to higher sell-through of loans to third parties on better terms. Gu pushed back on reading the take rate as a pricing lever, calling it "an output of the system" and naming contribution profit as the North Star.
Management estimates that without the 9-point rise in average UMI, originations would have been at least $370 million higher, with origination growth of 57% and contribution profit growth of 61%. That counterfactual is the core of the bull argument, but investors should note it is a counterfactual. October originations were about $1.25 billion, with daily volume down 9% from September and up 26% year-over-year. That is a sharp deceleration from the 44% Q3 growth rate. Blankmeyer said October pricing already reflects the September rate increase, because the marketplace reprices on roughly a monthly lag, and that spread improvements have partially offset the rise in Treasury yields. Management also said that on the unsecured side, with core borrowers often priced near the 36% rate cap, sensitivity to the cap is embedded in the 1% to 2% heuristic.
Secured Products Reach Profitability a Quarter Early, With Caveats
The most concrete operational result was secured contribution margin of positive 17%, versus negative 35% in Q2, a swing of more than 50 percentage points and a quarter ahead of the year-end target. Secured fee revenue was $34 million. Management cited three drivers: more loans sold to third-party capital partners on better terms, higher automation, and improved monetization. Nearly 40% of funded home loans were fully automated, up from under 5% a year ago, and fully automated auto-secured personal loans rose from 4% to 21%. In auto retail, rebuilding the credit application lifted end-to-end conversion by 31%.
The quality of the improvement deserves scrutiny. Secured fee revenue rose close to 50% sequentially, and Gu acknowledged "timing elements," saying some gains reflect things that "haven't been totally systematized" on loan sales and monetization. Volume was also sacrificed. Auto originations fell sequentially after rapid scaling, which Gu attributed "80%" to a deliberate focus on profitability and the remainder to macro. The company will hold secured contribution margin between breakeven and 25% and plans to return to growth in 2027. Investors should expect more quarter-to-quarter variability in this line than in unsecured, where contribution margin rose 2 points to 64%.
Upstart Bank Would Convert the Balance Sheet Into Cash
Management provided its most detailed outline yet of the planned bank launch, expected in early 2027 subject to remaining regulatory approvals. Upstart will capitalize the bank with about $404.1 million in cash and servicing assets. The bank will raise brokered deposits through two agreed partners, buy most of the existing loan book, and fund new originations. The equity currently tied up in that book will be released, and the related warehouse facilities will be paid down in an orderly way. Co-investments in committed capital partnerships will remain outside the bank. On the Q3 balance sheet, management said the transactions would produce pro forma consolidated unrestricted cash of more than $1 billion, with the bank's Tier 1 leverage and total risk-based capital ratios well above required minimums.
Management said capitalizing the bank requires no incremental equity. It also expects modest take rate and net interest income improvement because Upstart will keep fees and interest now paid to originating bank partners. Further benefits cited are a larger addressable market through underwriting all products in every state, lower deposit funding costs than warehouse funding, and a single regulatory regime in place of dozens to hundreds of bank partners. Management did not quantify margin impact. It reiterated that the model will still rely on third-party capital for the vast majority of loans, and that the 3-year guidance, including the 2028 target of a 25% adjusted EBITDA margin, already assumes the charter.
Capital Efficiency Is Improving, but From a Heavy Base
Loans held directly on the balance sheet fell to 4.8% of total loans outstanding, and more than 95% of Upstart-powered loans are third-party funded. About 68% of the platform is funded by committed capital partners, a level management considers about right and expects to hold roughly constant. Total equity used, covering balance sheet loans, co-investments and retained securitization interests net of financing, fell to 9% of platform loans from 11.6% a year ago. Platform loans grew 38% over the past year while equity used rose only 7%. Marginal equity use is lower still, with about $400 million of beneficial interest acquired year-to-date against just under $12 billion of originations, or about 3%. Management expects equity use to decline further as older co-investment vintages return cash and partnerships are renewed on better terms.
On credit, Blankmeyer said the last 12 quarterly vintages have returned about 660 basis points over Treasuries. Gu conceded that loans underwritten at lower UMI will see more delinquency in a higher UMI environment, but argued that model accuracy relative to UMI is what matters and that partner demand is the best evidence of attractive risk-adjusted returns. The company does not disclose vintage-level performance. Management put the sensitivity of fair value to a 100 basis point change in discount rates at about $16 million, meaning credit expectations matter far more than rates for the marks.
Model Improvements and Customer Growth
Upstart added more than 140 features to its core underwriting model in Q3, including proprietary signals on how consumers engage with the platform before borrowing. Management says its model is now 2.77x as accurate as a traditional credit scoring benchmark, with 87.27% of the inaccuracy gap still ahead. More than 1 million people applied for the first time, lifting cumulative unique applicants to over 25 million, while borrower acquisition costs fell 11% sequentially to $91 million. More than 555,000 borrowers took loans, a record, of whom over 250,000 were returning customers, a figure that combines repeat borrowing and cross-product activity. The company plans to discontinue its conversion rate disclosure, arguing the metric has become uninformative with multiple products, borrower segments and repeat customers, and will emphasize new versus returning borrower metrics instead. Some investors will view the removal of a long-standing metric with caution.
Gu also said inference and training costs are material for Upstart, unlike for traditional lenders, that these costs showed up in 2026 financials, and that efficiency work has progressed. The company uses cloud vendors and no on-premises servers.
Expenses and Operating Leverage
GAAP operating expenses were about $335 million, up 32% year-over-year and down 4% sequentially. Fixed expenses fell 3% to about $188 million, better than guided, though stock-based compensation of $41 million fell from $46 million on seasonality, and excluding it fixed costs were flat. Management guides to flat fixed expenses in Q4. Q3 net income was about $37 million, with diluted EPS of $0.33, and adjusted EBITDA was about $89 million, a 24% margin. The full-year guide implies a 19% margin and a Q4 margin in the high teens. Blankmeyer described 2026 as an investment year, with secured products and the bank absorbing spending, and pointed to improved operating leverage in 2027 and 2028 to support the 25% margin target. Full-year guidance now calls for fee revenue of about $1.32 billion and contribution profit of about $750 million, up roughly 41%.
Management's Case and Its Limits
Gu argued that the company does not need a macro tailwind, saying none of its forecasts depend on rates or UMI falling, and he noted that UMI has declined in exactly 15 of the last 30 quarters. He said a flat but elevated UMI is sufficient for multiyear growth, given the pace of alpha. The argument is plausible, but it rests on the fee business, where the guidance record is good. The balance sheet lines, which missed, do not benefit from most of the technology gains, and the company has only begun lowering their sensitivity. The bank and falling equity use are the credible path to doing so, though the benefit will arrive in 2027 at the earliest.
Gu was candid about valuation, saying he thinks investors are undervaluing the stock but "have every right to do that because they can't tell if and when we're finally going to make some real money." After 14 years, he said, the company has not produced meaningful profits yet. Given a full-year EBITDA guide cut by about 12%, a decelerating October, and unquantified downside to the fair value marks, that candor is warranted.