Upstart Holdings
This is the archived full research. Read the story
The company whose software decides who gets a personal loan, and at what rate, for more than 100 banks and credit unions.
The scores
Growth3/4 →
Revenue fell 39% in 2023, then grew 24%, 64% and 34% in Q3 2026, fees up 41%; the guide is cut and October slowed.
Business quality3/4 →
A data and model lead the regulator confirmed and a 64% core contribution margin, but Pagaya sells the same product and funders decide volume.
Financial position2/4 →
$1.6B of low-rate convertibles with no near maturity, but loans and loan securities it holds tie its capital to credit; funding is others' choice.
Execution record2/4 ↓
Profit returned in 2025, but the 2026 guide was cut, the 2025-2028 targets look out of reach, and the CEO and CFO are new.
Per share capture2/4 →
Stock pay of 11% of revenue is more than twice net income, and buybacks only partly offset it.
Potential2/4 →
A $293B pool of personal loan balances growing fast and secured loans just profitable, but Upstart can lend only what its buyers fund.
Risk profile2/4 ↓
One partner is 41% of revenue, profit swings with defaults, and the bank still needs two approvals.
Valuation4/4 →
About 2.13 times book, 50% of an excess-return value near $38: cheap on the base case, which itself needs a record return on equity.
Evidence quality3/4 →
Filed figures carry revenue, profit, funding and concentration, but the model's edge and the macro index rest mostly on management's word.
If it works 1.1× · if it fails you keep about 69%.
The Verdict
We rate Upstart Neutral, with low conviction. Its software approves more borrowers at lower rates than a credit score, a regulator found, and its fees grew 41% to a record. But it lends only what banks and investors choose to fund each month, one bank brings two fifths of its sales, and profit returned only last year after three years of losses. The price is about half our estimate of its value, so it looks cheap, but that estimate needs Upstart's best year ever, repeated while more borrowers fall behind. As a bet on car and home loans it is priced for success. Steady profit would change our view.
The long caseIf AI prices auto and home equity credit as well as it prices personal loans, Upstart's fee on each loan follows into much larger markets.
How we weighed it
The price against our value
In its favour
Price at 50% of our value
The price is about half our base value and close to our bear case, so much of the bad news is priced.
The business
Cuts both ways
Mixed
A model the regulator found approves more borrowers at lower rates, but funding and volume are decided by others.
Growth against what the price assumes
Cuts both ways
Fee revenue grows fast, but the outlook was cut and October slowed as defaults rose.
What could go wrong
Against it
One partner supplies two fifths of revenue, profit swings with defaults, and stock pay exceeds net income.
Against competitors
Cuts both ways
Pagaya sells the same model to lenders, and banks fund their own borrowers on deposits.
How its group is behaving
Against it
Consumer lenders trade on the credit cycle, and delinquencies on personal loans are rising across the industry.
The latest news and report
Cuts both ways
Record fees and a first profit on secured loans, against a cut guide and a new chief accounting officer to find.
Where this departs from the fair value
The base case needs the return on equity to climb to its best filed year and hold through a cycle that once brought years of losses. Funders and defaults decide that, both are moving against it, so the discount is owed for that risk.
The evidence under this - 7 claims
The return Upstart earned turned to losses in 2022-2024 (C04); if funding pulls back again the bear case near $13 is the closer reading.
The index is Upstart's own and unpublished; TransUnion's count of personal loans 60 days late rose to 4.12% from 3.52% (C31), which points the same way from outside.
More than half of funding runs through committed partnerships in which Upstart co-invests (S01), so its capital still rides on the loans it does not hold.
The company
Upstart, based in California and founded in 2012, built software that looks at more than a credit score: job, education, bank data and how people use the site. It approves more than 90% of loans with no person involved.
More than 100 banks and credit unions use it to find borrowers. Beyond personal loans it now offers car loans, home equity loans and a credit line.
The evidence under this - 3 claims
No outside party has tested the 2.77 figure; the last regulator reading, the CFPB's, dates from 2019 (C16).
How it makes money
Nearly nine tenths of sales are fees on personal loans, paid by the lender. About a tenth is the same fee on car and home equity loans. A small part is interest on loans Upstart keeps, less their losses. One lending partner alone brought 41% of sales in the third quarter.
The evidence under this - 4 claims
Where the money comes from, stream by stream
| Stream | Share of sales (Q3 2026) | How it earns it |
|---|---|---|
| Fees from unsecured loans | 89% | Referral, platform and servicing fees on personal loans, small-dollar loans and Cash Line, paid by lenders and investors. |
| Fees from secured loans | 9% | The same fees on auto loans and home equity lines. |
| Interest and fair value on loans held | 2% | Interest on loans and securities Upstart holds, less the cost of funding them and changes in their value. |
Who runs it
Paul Gu, a co-founder who built the software, became chief executive in May 2026; the other co-founder, Dave Girouard, stays on the board and owns 13.51%. Gu has bought shares with his own money three times this year. The finance chief is new, and the chief accounting officer leaves in November.
The evidence under this - 3 claims
Why it wins, and who could beat it
Why customers come back
Every loan repaid or missed teaches the software, and Upstart has nearly 104 million repayments to learn from. In 2019 the US consumer finance regulator found its model approved 27% more people than a traditional one, at 16% lower rates.
That edge shows in what it keeps: on personal loans, 64 cents of each fee dollar is left after finding and serving borrowers.
The evidence under this - 4 claims
No outside party has tested the 2.77 figure; the last regulator reading, the CFPB's, dates from 2019 (C16).
Fair value write-downs as the macro index rose (C07) say recent loans are performing worse than priced.
The threat
Pagaya sells banks a similar AI model. SoFi, a bank, lends to the same borrowers from its own deposits, and card issuers like Capital One hold the debt a personal loan would replace. A big bank could also build its own model.
The evidence under this - 3 claims
The rivals: PGY, SOFI, COF, JPM, ALLY, FIGR, and 1 more that is private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| PGY | PagayaPriced against it: Not checked | Unsecured lending marketplace | On sale | Not checked |
| SOFI | SoFiPriced against it: Not checked | Unsecured lending marketplace | On sale | Not checked |
| COF · JPM | Credit card issuers such as Capital One and JPMorgan ChasePriced against it: Not checked | Unsecured lending marketplace | On sale | Not checked |
| In-house | Lenders' own credit-score modelsPriced against it: In-house | Unsecured lending marketplace | On sale | In-house |
| ALLY · FIGR | Ally and Figure in auto loans and home equity linesPriced against it: Not checked | Secured lending: auto loans and home equity lines | On sale | Not checked |
How big it could get
How big this could get
Americans owe about $293B on personal loans, up 9.0% in a year, and take out about 8.4 million new ones a quarter, a record, says TransUnion, a credit bureau. Upstart funded 574,089 loans in its latest quarter, so most of that market is still open to it.
The evidence under this - 2 claims
The results
The numbers
Sales grew 39% over the twelve months to Sep 2026, slower than the 64% in the year to Dec 2025.
Operating margin is 4.6%, up from −50% in the year to Dec 2023.
Yearly sales growth has ranged from −39% to +264% over the last 5 years, with sales falling in 2 of them.
Shareholders' money earned 8.0% a year, down from 24% in the year to Dec 2021.
It used $187M of cash over the last twelve months, so the $461M it holds lasts about 2.5 years at that pace.
It owes $1.57B more than the cash it holds, and its operations bring in no cash to repay it.
The share count rose 32% in 4.8 years, and pay in shares costs 11% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-11-17. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Sep 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $849M | $842M | $514M | $637M | $1.04B | $1.34B |
| Sales growth | +264% | −0.7% | −39% | +24% | +64% | +39% |
| Profit | ||||||
| Operating margin | 17% | −14% | −50% | −27% | 4.1% | 4.6% |
| Net margin | 16% | −13% | −47% | −20% | 5.1% | 4.9% |
| What growth costs, per sales dollar | ||||||
| Research and development | 16% | 28% | 55% | 40% | 25% | 25% |
| Sales and marketing | 39% | 41% | 25% | 26% | 29% | 31% |
| General and admin | 14% | 22% | 41% | 36% | 24% | 22% |
| Pay in shares | 8.6% | 15% | 34% | 21% | 13% | 11% |
| Capital spending | 1.0% | 1.0% | 0.3% | 0.1% | 0.0% | 0.4% |
| Cash | ||||||
| Cash from operations | $168M | −$658M | −$112M | $186M | −$148M | −$181M |
| Free cash flow | $160M | −$667M | −$113M | $185M | −$148M | −$187M |
| Balance sheet | ||||||
| Cash and short-term investments | - | - | $368M | $788M | $652M | $461M |
| Debt | $695M | $986M | $1.04B | $1.40B | $1.83B | $2.03B |
| Net cash | - | - | −$672M | −$614M | −$1.18B | −$1.57B |
| Returns and shares | ||||||
| Shares (diluted) | 94.8M | 82.8M | 83.8M | 89.5M | 107M | 125M |
What growth costs
Growth is getting cheaper per dollar of sales: each employee now brings in nearly twice what one did in 2020, and Upstart turned a profit. But it pays staff about 11 cents of each sales dollar in shares, more than it earns.
The evidence under this - 4 claims
The last report (3 Nov)
Quarter ended September 2026, filed figures.
In the third quarter of 2026 loans arranged rose 44% to $4.1B and fees rose 41% to a record $364M. Profit under accounting rules was $36.6M. Car and home equity loans made money for the first time.
But the interest earned on loans Upstart holds fell, because more borrowers are falling behind, and it cut its 2026 sales forecast to about $1.35B. October lending slowed.
What it did to the thesisIt showed the core lending business working, with record fees and a first profit on car and home loans, but rising late payments hit the loans it holds, which is the risk the case turns on.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.33 against $0.20 expected; $0.33 as filed, with no adjustments.
The evidence under this - 6 claims
October originations of about $1.25B, with daily volume down 9% on September (C08), already point below the third quarter's pace.
What management did with the money
Upstart bought back $100.1M of shares this year, paid off notes due in 2026, and borrowed cheaply through notes that turn into shares only far above today's price. It is putting about $404M into its own bank, which should let it fund loans with deposits.
Insiders in 2026Q3: 2 purchases and 11 sales, from their Form 4 filings.
The evidence under this - 3 claims
Two of three approvals are outstanding, and the cash figure is management's illustration on the Q3 balance sheet.
The price, and where we disagree
In short. Cheap. The price is about 50% of what we think the business is worth, at an 8% yearly return.
The shares fell 48.8% in a year, to a 52-week low of $19.09, because buyers pay far less for each dollar of sales even as sales grew. They rose after the third-quarter report, then fell again.
The next check is the fourth-quarter report in February 2027, with a forecast for 2027. Upstart's own bank could open early in 2027 if two more regulators approve it.
The evidence under this - 4 claims
Two of three approvals are outstanding, and the cash figure is management's illustration on the Q3 balance sheet.
October originations of about $1.25B, with daily volume down 9% on September (C08), already point below the third quarter's pace.
How we value it
- Method
- The owners' book plus the returns it earns above the rate
- Required return
- 8%
- Long-run growth
- 4%
- Moat
- An average moat
| Bear | Base | Bull | |
|---|---|---|---|
| Book growth, years 1 to 3 | 11.3%, 13.8%, 15% | 22.7%, 27.6%, 30% | 28.4%, 34.5%, 37.5% |
| Normal return on equity | 13.4% | 16.8% | 18.5% |
| Return on new capital after year ten | 8% | 12% | 15% |
| Value a share | $13.05 | $37.87 | $67.34 |
- Price
- $19.09
- Priced on
- 16 Nov
- Price as a share of the base value
- 50%
- Yearly return the price offers on the base case
- 10.4%
- Risk check: valued at 10%, the rate for a cyclical or indebted business
- $20.95
Where each figure comes from
| Figure | Source | Where exactly |
|---|---|---|
| The owners' book | Filed report | Quarterly report (10-Q) for the period to 30 Sep 2026, filed 3 Nov 2026; the balance sheet on that date. |
| Shares | Filed report | Quarterly report (10-Q) for the period to 30 Sep 2026, filed 3 Nov 2026; the diluted share count. |
| Book growth, years 1 to 3 | Analysts' forecasts | FMP's consensus for the years to 31 Dec 2026, 31 Dec 2027 and 31 Dec 2028: the average of 11 analysts' revenue estimates; the owners' book grows on that path, held to 30% a year. FMP publishes how many analysts, not who they are. |
| Normal return on equity | Our judgement | Our reading of the evidence; each case below names it. |
| Return on new capital after year ten | Our method's rule | Set by our reading of the moat: the 8% rate with none, 12% with an average one, 15% with a strong one; the bear case earns the rate. |
| The bear and bull cases | Our method's rule | Our method's spread: the bear case takes half the growth and a fifth off the margin, the bull a quarter more growth and a tenth more margin. |
| Price | Market price | FMP's closing price on 16 Nov 2026. |
What each case rests on
Bear caseEquity grows at half the analysts' path, 11%, 14% and 15%, the return on it settles at 13.4%, a fifth under the base, and new equity earns only the 8% rate: the test if funding pulls back as in 2022.
Base caseEquity grows 23%, 28% and 30%, the analysts' revenue path (C28) held to 30% in year three, then fades toward 4%; the return on equity rises to 16.8%, its best filed year (C42), below the analysts' 2028 forecast near 20%.
Bull caseEquity grows a quarter faster, 28%, 35% and 38%, the return on equity reaches 18.5%, and new equity earns 15%, as if the bank lowers funding costs and secured loans add a second leg.
What moves the base value
| Lower | Higher | |
|---|---|---|
| Growth in years 1 to 3, a quarter lower or higher | $30.94 | $46.64 |
| The normal return on equity, 15% lower or higher | $34.74 | $41.00 |
| Return on new capital after year ten, at the rate or one step up | $19.75 | $51.46 |
| The rate, a point higher or lower | $27.65 | $55.13 |
| Long-run growth a point lower | $33.45 | - |
Value a share as growth and margin change
| Normal return on equity ↓ · growth → | 50% | 75% | 100% | 125% | 150% |
|---|---|---|---|---|---|
| 13.4% | $22.91 | $27.64 | $33.70 | $41.40 | $51.10 |
| 15.1% | $24.21 | $29.29 | $35.79 | $44.02 | $54.37 |
| 16.8% | $25.51 | $30.94 | $37.87 | $46.64 | $57.63 |
| 18.5% | $26.81 | $32.59 | $39.95 | $49.25 | $60.90 |
| 20.1% | $28.10 | $34.24 | $42.04 | $51.87 | $64.16 |
Growth is shown as a share of the base case's. Green: the price is at or under 80% of that value. Red: at or over 120%.
How we read the multiple
The shares are worth about $1.9B, about 2.1 times what Upstart owns minus what it owes, and about 29 times its profit over the last four quarters.
Analysts expect sales to grow about 29% a year to 2028. On that path, with profit climbing to Upstart's best year as a share of what it owns, we value the shares at about $38, $13 in our gloomy case and $67 in our sunny one.
Against its own history. Buyers now pay far less for each dollar of sales than a year ago, while sales per share grew.
The evidence under this - 4 claims
On the 125.4M diluted shares that count the notes as converted (C13) the market value is higher, but the notes convert far above the price (C12).
The return Upstart earned turned to losses in 2022-2024 (C04); if funding pulls back again the bear case near $13 is the closer reading.
The 2027 figure needs revenue to grow 29% after a guide cut, with the macro index rising (C06, C08).
Priced against its rivals
None of the listed rivals above sells the same thing to the same buyers as a large part of what it does. A price comparison needs at least two, so the price is judged against its own history instead.
The bet: what it pays if it works
The bet is on car loans and home equity loans. Loans arranged there nearly tripled in a year and just turned a profit, but they bring about $34M a quarter in fees.
If they grow to $400M a year, the shares would be worth only about a tenth more than today, because the price already counts on that growth. So this bet is priced for success.
| Secured lending: auto loans and home equity lines revenue in two to three years | Dollars of secured loans originated a year needed | At the stated multiple |
|---|---|---|
| $200.00M | about 3.4B at 100% kept | 0.9× |
| $400.00M (the big win) | about 6.9B at 100% kept | 1.1× |
| $800.00M | about 13.8B at 100% kept | 1.5× |
To pay 2× (a Fair bet) it would need about $1.24B of Secured lending: auto loans and home equity lines revenue (about 21.3B dollars of secured loans originated a year); to pay 5× (a Cheap bet), about $4.07B of Secured lending: auto loans and home equity lines revenue (about 70.2B dollars of secured loans originated a year), at the same multiple.
Each row is Secured lending: auto loans and home equity lines alone, plus the rest of the company (Unsecured lending marketplace: personal loans, small-dollar loans and Cash Line) counted at about $1.30B.
The steps it has to take
- Done Auto and home equity lending launched with partners
- Done Secured contribution margin positive
- In progress Secured loans funded by credit-union partners at scale due 2027-12
- Not started Secured segments profitable for a full year due 2027-12
The detail behind the label
The evidence under this - 3 claims
On the 125.4M diluted shares that count the notes as converted (C13) the market value is higher, but the notes convert far above the price (C12).
Where we disagree with the market
A lender that rises and falls with the economy, whose AI has not yet made steady money.
The AI is a real edge and the price is low, but steady profit depends on funders and late payments Upstart cannot control.
Analysts are split evenly, and nearly a third of the shares that trade are bet against it, while the chief executive buys. We agree the shares look cheap, but our value needs a record year that lasts.
Where we could be wrong: if Upstart earns 12% a year on what it owns for four quarters while late payments stay high, the cheap price would be the real story.
What analysts say, and the evidence
The evidence under this - 5 claims
The return Upstart earned turned to losses in 2022-2024 (C04); if funding pulls back again the bear case near $13 is the closer reading.
What could go wrong, and how we would know
The risks
- One big partner. A single bank brought 41% of sales last quarter and could move its lending.
- Others decide the money. Banks and investors choose each month how many loans to fund; in 2022 they pulled back and sales fell.
- Late payments. As more borrowers fall behind, Upstart approves fewer loans and the loans it holds lose value.
- Shares paid to staff. Upstart pays staff more in shares than it earns in profit.
- The bank may wait. Two regulators still have to approve its bank.
The evidence under this - 5 claims
More than half of funding runs through committed partnerships in which Upstart co-invests (S01), so its capital still rides on the loans it does not hold.
The index is Upstart's own and unpublished; TransUnion's count of personal loans 60 days late rose to 4.12% from 3.52% (C31), which points the same way from outside.
Two of three approvals are outstanding, and the cash figure is management's illustration on the Q3 balance sheet.
What we still do not know
- How Upstart's loans perform against ordinary ones to the same borrowers as late payments rise; nobody outside has checked since 2019.
- Who the biggest lending partner is, and how long it has agreed to stay.
- When the last two regulators will rule on its bank.
The evidence under this - 3 claims
No outside party has tested the 2.77 figure; the last regulator reading, the CFPB's, dates from 2019 (C16).
Two of three approvals are outstanding, and the cash figure is management's illustration on the Q3 balance sheet.
What would change our view
- Neutral→AttractiveProfit proves the return
It needs:
- ○Upstart earns 12% a year on what it owns for four quarters while late payments stay high.checked at the next results, due by 9 Feb
- Neutral→UnattractiveThe price runs 20% above fair value
It needs:
- ○The share price rises to about $45.44, 20% above what we think the business is worth.about $45.35 today, 138% above the last close
- Neutral→UnattractiveFunding pulls back
It needs any of:
- ○Loans arranged fall from a year earlier for two quarters running.checked at the next results, due by 9 Feb
- ○A lending partner bringing a tenth or more of sales ends or shrinks its lending.watched as it happens
We would also be wrong if
- Profit kept on personal loans falls below 55 cents of each fee dollar for two quarters.
- Its bank opens and funds the loans it holds with deposits.
What is coming
- 9 February 2027 Q4 2026 results and the 2027 outlook: Whether originations grow again after October's dip and what return on equity the 2027 guide implies with the macro index where it is.
Every dated event ahead
- 9 February 2027 Q4 2026 results and the 2027 outlook: Whether originations grow again after October's dip and what return on equity the 2027 guide implies with the macro index where it is.
- 31 March 2027 Upstart Bank opens: Whether deposits fund the held loans, cutting funding costs and freeing over $1B of cash, as management says.
- 30 April 2027 2027 proxy statement: Whether pay keeps riding on fee revenue, and the founders' holdings.
Since we wrote this
- 17 NovThesis writtenNeutral, low convictionThe price is about half our value, but that value needs a record return on equity through a credit cycle that funders and defaults decide.
Sources 15
Sources - 15
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-02-10Item 1 products, the model's variables and repayment events, the internal approval study, the funding mix, the top three lending partners, the competition classes, loan terms and APR caps, headcount, statementsbacksC01C03C04C09C10C14C15C17C34C36C39
- S02primarycompany filing2026-11-03customer concentration, the balance sheet, the convertible notes and warehouse lines, the share count, the buyback, the segment note and the contribution of the other segmentsbacksC02C03C04C09C11C12C13C14C18
- S03primarycompany filing2026-11-03originations, revenue, fee revenue by product, contribution profit and margin, GAAP and adjusted profit, and the cut full-year 2026 outlookbacksC01C02C05C06C07C18C19C38C41
- S04primarytranscript2026-11-03Q3 2026 earnings call transcript (FMP)data/cache/transcripts/UPST-2026Q3.jsonthe Upstart Macro Index and its effect on originations, secured originations, the take rate, funding and equity use, vintage returns, Upstart Bank's funding plan, October originations and the chief executive's answer on profits
Our model is now 2.77x as accurate, and 87.27% of the inaccuracy gap is still ahead of us.
backsC06C07C08C10C15C18C20C21C40 - S05primarycompany filing2026-02-102025 originations and revenue, the 2025-2028 targets and the chief executive transitionbacksC22C35
- S06primarycompany filing2026-04-16executive officers, beneficial ownership, bonus and performance-share metricsbacksC22C23
- S07primarycompany filing2026-11-09the chief accounting officer's resignation, effective 30 November 2026backsC22
- S08primarycompany filing2026-10-20a new $200M revolving credit agreement signed on 16 October 2026backsC12
- S09primaryregulator2019-08-06CFPB blog: An update on credit access and the Bureau's first No-Action Letter Consumer Financial Protection Bureauthe regulator's reading of Upstart's model against a traditional one, from data Upstart supplied under its no-action letter
the tested model approves 27% more applicants than the traditional model, and yields 16% lower average APRs for approved loans
backsC16 - S10primaryindustry data2026-11-05US unsecured personal loan originations, balances and serious delinquency
originations climbing 22.7% YoY to a record 8.41 million loans and outstanding balances rising 9.0% to $293.4 billion
backsC31 - S11primarycompany filing2026-03-02a rival's description of the same product: AI underwriting connecting lenders with investors, in personal, auto and point-of-sale loans
a network connecting financial institutions and their customers with investors
backsC32 - S12primarycompany filing2026-02-17a deposit-funded lender's personal loan growth in 2025
During the year ended December 31, 2025, total personal loan origination volume increased by 56% relative to 2024
backsC33 - S13secondaryvendor data2026-11-17UPST research pack (FMP market tier, FINRA short interest, headcount), frozen 2026-11-17content/UPST/research-pack.jsonthe 16 November close of $19.09, daily bars, returns and their split, analyst ratings and their history, earnings history, insider filings, short interest, headcountbacksC24C25C26C27C36C37C38
- S14secondaryvendor data2026-11-17the analysts' average revenue and net income for 2026 to 2029, with the count of analystsbacksC28
- S15primarycompany filing2026-11-17every open-market purchase and sale, and its Rule 10b5-1 flagbacksC24