Share count
107Mpeak FY2025
← All companies ← UPS Manual UPWK → ← TW Capital Markets & Asset Management UROY →
UPST, Upstart
Capital Markets & Asset Management financial
Upstart is the leading artificial intelligence lending marketplace.
Long-term, our vision is to become the always-on, everything-store for credit, where we can automatically approve borrowers at the right prices instantly and effortlessly.
While FICO is rarely the only input in a lending decision, most lenders use simple rules-based systems that consider only a limited number of variables.
The business
What it sells, where the money comes from, the kind of company it is.
Revenue · FY2025
$1.0B
+64.0% YoY · 35% 5-yr CAGR
Vital signs · TTM, with 5-yr average
Revenue $1.1B 5-yr avg $780M
Revenue $308M +44.4% vs yr-ago qtr
Operating margin −2.4% yr-ago −2.1%
Net income ($7M) yr-ago ($2M)
Net margin −2.2% yr-ago −1.1%
The business in brief
read the 10-K →What this business is and what moves its needle, from its own SEC filings.
- What moves the needle
- Net interest margin, loan losses, and book value. A lender is read on the quality of its balance sheet, not an earnings multiple, and the worst year of credit losses matters more than the best. On its own account, the filing leans hardest on customer concentration, set against the numbers in what the filing emphasizes, below.
- Is it a good business?
- Return on equity has sat below the cost of equity (median 1%, above 12% in only 1 of 7 years). The cycle and the loan book decide this one; weigh the recession years in the record, not the average, and read the 10-K.
Every line is arithmetic on the company's filings, shown in full in the sections below.
The record
Ten years of arithmetic, read across the cycle.
The record, 2019–2025
realized figures from each filing · older years to the left| 2019’19 | 2020’20 | 2021’21 | 2022’22 | 2023’23 | 2024’24 | 2025’25 | TTMTTMMar 2026 | |
|---|---|---|---|---|---|---|---|---|
| Income statement | ||||||||
| $160M | $229M | $801M | $907M | $514M | $637M | $1.0B | $1.1B | RevenueRevenue |
| — | — | $47M | ($65M) | ($47M) | $1M | $94M | $97M | Net interest incomeNet int. |
| ($466K) | $6M | $135M | ($109M) | ($240M) | ($129M) | $54M | $49M | Net incomeNet inc. |
| — | 6% | -1% | — | — | — | 1% | 2% | Effective tax rateTax rate |
| Cash flow & returns | ||||||||
| — | 1.3% | 7.4% | -5.6% | -11.9% | -5.4% | 1.8% | 1.7% | Return on assetsROA |
| — | 2% | 17% | -16% | -38% | -20% | 7% | 7% | Return on equityROE |
| — | 2% | 17% | −16% | −38% | −20% | 7% | 7% | Retained to equityRetained/eq |
| — | 2% | 19% | -18% | -43% | -23% | 7% | 8% | Return on tangible equityROTCE |
| Balance sheet | ||||||||
| — | $477M | $1.8B | $1.9B | $2.0B | $2.4B | $3.0B | $3.0B | Total assetsAssets |
| — | $0 | $67M | $67M | $67M | $67M | $67M | $67M | GoodwillGoodwill |
| ($62M) | $300M | $807M | $672M | $635M | $633M | $799M | $733M | Shareholders’ equityEquity |
| Per share | ||||||||
| 14.3M | 17.5M | 94.8M | 82.8M | 83.8M | 89.5M | 107M | 96.9M | Shares out (diluted)Shares |
| $-0.03 | $0.34 | $1.43 | $-1.31 | $-2.87 | $-1.44 | $0.50 | $0.51 | EPS (diluted)EPS |
| $-4.30 | $17.14 | $8.52 | $8.12 | $7.58 | $7.08 | $7.43 | $7.57 | Book value / shareBVPS |
| $-4.30 | $17.14 | $7.60 | $7.12 | $6.65 | $6.22 | $6.73 | $6.79 | Tangible book / shareTBVPS |
The diluted share count moved ×5.41 into 2021 — shares issued, not a split the totals corroborate — and the per-share figures carry the counts as filed.
Per-share growththe realized rate an owner's share compounded
| 6-yr | 5-yr | |
|---|---|---|
| Revenue / share | −2.3%/yr | −5.7%/yr |
| EPS | — | +7.9%/yr |
| Capital spending / share | −52.5%/yr | −47.0%/yr |
| Book value / share | — | −15.4%/yr |
The record, charted
FY2019–2025Each measure over its full record; the current point and the worst year marked.
Revenue
$1.0Blow FY2019
Quality & stewardship
Returns, the balance sheet, capital allocation, and pay.
Owner’s Scorecard
FY2025 10-K · source on SEC EDGAR →
Is it a good business?
- Below the cost of equityNet income $54M ÷ equity $799MIndustry peers: median 5%
What this means
The bank's north star, what it earns on shareholders' capital. Cost of equity is roughly 10%, so a return durably above that builds value and below it destroys it. One year is noisy; the durability across a full credit cycle is what counts.
- ModestNet income ÷ (equity − goodwill $67M − intangibles $8M)Industry peers: median 2%
What this means
The cleaner return, stripping out the goodwill paid for past acquisitions. This is the number a buyer of the whole bank actually earns on the hard capital.
- Not enough data
What this means
Noninterest expense or revenue missing.
Is it sound?
- Capital (equity / assets) 26.9%Well capitalizedEquity $799M ÷ assets $3.0B
What this means
A plain-English leverage read: how much of the balance sheet is the owners' own money. This is a rough proxy; the regulatory figure is the CET1 ratio, which is risk-weighted and reported in the filing. The point is the same, how much loss the bank can absorb before depositors are at risk.
- Funding —Not enough data
What this means
Deposits or total assets missing.
- Credit cost —Not enough data
What this means
Provision or net interest income missing.
Does AI threaten the moat?
Low contestabilityThe moat is physical, regulated or balance-sheet-funded, the kind AI cuts costs within but does not contest.
In its own filing A competitive risk, new this year
Its FY2025 10-K names artificial intelligence as a competitive threat, in language that was not in the prior year's filing.
“Any significant disruption of, or failure in, our technology systems, including our AI lending platform, could adversely affect our business, financial condition and results of operations.”
AI is unlikely to contest a moat that is physical, regulated or balance-sheet-funded; here it reads more as a cost tool than a threat, and the company is using it that way.
Read from the filing's own risk factors, paired with the industry's structure under its SIC code; the durability is read above, the price below.
All figures as filed; the source filing is linked above.
Management, ownership & pay
read the proxy →From the proxy: how much of the business the people running it own, and how they are paid, beside what the business earned for its owners in the same years.
| Fiscal year | Chief executive | Pay, as filed | “Actually paid” | Owner earnings |
|---|---|---|---|---|
| 2021 | Dave Girouard | $15.0M | $70.4M | $160M |
| 2022 | Dave Girouard | $9.6M | −$36.2M | ($667M) |
| 2023 | Dave Girouard | $7.1M | $26.1M | ($113M) |
| 2024 | Dave Girouard | $7.1M | $26.7M | $185M |
| 2025 | Dave Girouard | $9.7M | $855k | ($148M) |
Both pay figures are the company’s own, from the pay-versus-performance table its proxy statement files. “As filed” is the Summary Compensation Table total: salary, bonus, and equity awards at their value on the day of grant. “Actually paid” is the SEC’s prescribed recalculation, which re-marks those equity awards to what they became as they vested; it can swing far above or below the filed figure in either direction, and negative years occur. Owner earnings are the whole business's, from the record above, for the same fiscal years.
- Insider ownership17.3%
The stake all directors and executive officers hold together, per the 2026 proxy: skin in the game, the first thing Munger reads.
- CEO pay ratio34:1
What the chief earns for every dollar the median employee makes, per the 2026 proxy. A high ratio alone settles nothing; some businesses are genuinely top-heavy in scarce skill. A runaway figure is where Buffett starts asking whether the board is doing its job.
- Stock-based compensation$132M
The slice of the business handed to employees in shares this year, 13% of revenue, equal to 310% of operating profit. Buffett's oldest accounting fight: this is compensation, compensation is an expense, real whether or not the headline earnings admit it. One trap: the cash-flow statement adds SBC back, so the operating cash, and the owner earnings drawn from it, are flattered by exactly this amount; counted as the cost it is, what an owner keeps is lower.
Peers, Capital Markets & Asset Management
The same industry, side by side on the bank lens. Each figure is a through-cycle median, so a peak or trough year can’t distort it; the group median at the foot is the line to read each against.
| Company | Revenue | ROE | ROTCE | Efficiency | NII / assets |
|---|---|---|---|---|---|
| BKKTBakkt Inc. | $2.3B | -146% | -252% | — | 0.5% |
| GDOTGreen DOT Corp | $2.0B | 5% | 13% | — | -0.1% |
| ATLCAtlanticus Holdings Corporation | $2.0B | 42% | 42% | 55% | 21.3% |
| TREELendingTree Inc. | $1.1B | 5% | — | — | -2.2% |
| UPSTUpstart | $1.0B | -7% | -8% | — | 0.0% |
| SOFISoFi Technologies | $3.6B | -6% | -9% | 85% | 3.6% |
| ABTCAmerican Bitcoin Corp. | $185M | -38% | -152% | — | 0.0% |
| VELVelocity Financial Inc. | $186M | 16% | 16% | — | 2.5% |
| Group median | — | -0% | -8% | — | 0.3% |
The price
What a price has to assume.
What the price implies
price / tangible bookA bank is worth a multiple of its tangible book value, and the multiple it deserves is set by the return it earns on that book. Type today’s price; we show what you would be paying against what Upstart’s record justifies.
$
The assumptions
Tangible book / share, delivered−14%/yr’20→’25
The justified multiple is (return on tangible equity − growth) ÷ (cost of equity − growth). A bank earning exactly its cost of equity is worth about one times tangible book; the premium above that prices each point of durable excess return. A higher cost of equity lowers the justified multiple for a bank.
Enter a price above to run it.
Price / tangible book—
Justified by the return—
Normalized return on tangible equity−8%
Price / book—
Earnings yield—
P/E (3-yr avg ’23–’25)—
Graham’s price gate—
Graham applied the same standards to financial enterprises (Intelligent Investor ch.14): the 15× multiple cap on averaged earnings, and P/E times price-to-book at most 22.5. The gate marks the bargain-hunter’s floor, not a verdict.
A dated snapshot of the price you typed, the assumptions you set, and what the page showed for them. A snapshot is never edited after it is saved. Your notebook is yours alone — the commitment states what is stored and what we will never do.
Tangible book $658M on 96M shares, a −8% normalized return on it. The dials set the multiple such a return would justify; your price sets the multiple you are paying. It assumes the bank keeps earning that return; a credit cycle, a rate shock or a bad acquisition changes it, which is what the record and the 10-K are for.
Cite: Owner Scorecard, "Upstart (UPST), the owner's record," https://ownerscorecard.com/c/UPST, data as of 2026-07-09.
Manual order: ← UPS its page in the Manual UPWK →
Industry order: ← TW the Capital Markets & Asset Management chapter UROY →