I opened Affirm's annual report with one narrow question. I wanted to see the capital structure of a large consumer lender that has no deposits at all, and Affirm is the cleanest public example I know.
The report was filed on 27 August 2026. Total assets were $15.79bn, total liabilities $10.31bn and stockholders' equity $5.48bn. A year earlier those lines read $11.15bn, $8.09bn and $3.07bn.
Equity rose by 79% in twelve months. That number took me an afternoon to explain. The explanation was not the one I expected.
If you read any lender's balance sheet, you should start with the funding side before the loan book. A bank answers that question with one word, deposits. Affirm needs a whole page of the filing to answer it.
Affirm says its funding model "consists of four primary channels". It lists warehouse credit lines, term and revolving securitizations, pass-through loan sales and forward flow sales to outside buyers. I went looking for those channels on the balance sheet. I found two of them.
"Notes issued by securitization trusts" stood at $5.33bn on 30 June 2026. "Funding debt", which the filing ties mainly to warehouses, stood at $3.33bn, up from $1.62bn a year before. Together that is $8.66bn of secured debt, 84% of everything Affirm owes.
Funding debt more than doubled in one year. Loans held for investment grew 36%, to $9.56bn. So the warehouse borrowing grew roughly three times as fast as the book it finances.
This debt also turns over fast. During the year Affirm received "$40.9 billion in proceeds from the issuance of secured debt" and repaid $38.7bn. I read that as the same few billion dollars being borrowed, used and returned several times a year, which is what the company means when it writes that "funding can be recycled quickly".
I keep one piece of research next to this filing. It is NBER working paper 26903, first written in 2019 as "Banking without Deposits" and revised in June 2024, by Erica Jiang, Gregor Matvos, Tomasz Piskorski and Amit Seru.
Their starting point is the ordinary bank. They describe banks as firms with "debt financing accounting for about 90% of banks' assets". That leaves roughly 10% for equity.
Their main finding concerns lenders like Affirm. "We demonstrate that shadow banks employ twice the amount of equity capital compared to equivalent banks." They also estimate that without deposit insurance, smaller and mid-size banks "would maintain a capitalization level at least 25% higher in relative terms".
I had assumed a fintech lender would run thinner than a bank. The paper says the opposite. Affirm agrees with the paper.
Its equity was 34.7% of total assets in June 2026 and 27.5% a year earlier. Both are far above the bank benchmark the paper uses. Without insured deposits, the market makes a lender like this hold more of its own money, and I think that is the single most useful idea for anyone pricing a fintech lender.
Affirm also reports its own capital metric. "As of both June 30, 2026 and June 30, 2025, our equity capital as a percentage of our total platform portfolio was 4%."
The platform portfolio includes loans Affirm has sold to third parties. So the denominator is bigger than the balance sheet. I have not found the year-end portfolio figure in the filing text, only the average, which rose by $4.6bn, or 35%.
Here is what confuses me. Book equity rose 79%, yet the ratio stayed at 4% in both years. Either "equity capital" in that sentence is narrower than book equity, or the rounding hides a large move. I cannot tell which from the text I read.
The answer to the 79% sits in one line. Deferred tax assets were $13.9m in June 2025. In June 2026 they were $1.47bn.
Affirm explains it plainly. It "released a significant portion of our domestic valuation allowance , resulting in a non-cash income tax benefit of approximately $1.5 billion during the year ended June 30, 2026." Net income for the year was $1.93bn, against $52.2m the year before.
The company gave weight to "our achievement of a cumulative U.S. income position over the three-year period". Three years of cumulative U.S. income let the accountants say that old losses will now save real tax.
Most of the equity increase is therefore not new money. Equity rose by $2.41bn and the tax asset by $1.45bn, which is 60% of the change. Option and warrant exercises and the employee share plan brought in only $158.9m of cash.
I stripped the tax asset out of both sides. Equity drops to $4.02bn and equity to assets falls to 28.0%. Liabilities per dollar of equity become 2.57, against 1.88 reported and 2.63 a year earlier.
On that view the capital structure barely moved. I find that hard to read past.
A deferred tax asset cannot repay a warehouse line. It has value only while Affirm keeps earning taxable income in the United States, and that is the condition I would expect to fail first in a bad credit year.
Convertible notes are the other layer. On 30 June 2026 Affirm owed $221.3m of "0.00% convertible senior notes due November 15, 2026". It also owed $920.0m of 0.75% notes due December 2029.
Those 2029 notes were sold in the previous fiscal year. In that year Affirm spent $1.0bn retiring part of the 2026 notes and $250.0m on "the repurchase of shares of our Class A common stock in connection with the issuance of the 2029 Notes".
I like that sequence as a piece of capital structure management. Near-dated notes were swapped for later ones, and shares were bought back alongside the new issue.
The 2029 notes carry a book value of $908.5m. Their fair value is $1.1bn. My guess is that the conversion option explains most of that gap, which would make this debt behave partly like equity.
This has nothing to do with the funding mix. GMV was $50.2bn in the year, up 37%, and "GMV attributable to Amazon represented 22% of total GMV", the same share as in 2025.
One merchant drives over a fifth of volume. I keep thinking about that whenever I look at the funding lines. I cannot connect the two with anything in the filing.
Back to the ratios.
The part of Affirm's capital structure that I worry about most sits off the balance sheet. Forward flow buyers, including private credit funds, commit to buy loans for terms "ranging between one to three years".
Affirm is candid about them. "One or more of these counterparties have experienced redemption pressures in the past." If buyers pull back, the company says it may need to "retain more loans on our balance sheet".
That would push more weight onto its spare capacity. At year end Affirm reported $5.3bn of unused funding debt capacity, $2.6bn of cash and securities and $675.0m of undrawn revolving credit.
It is a sizeable cushion. It is also a cushion for a company whose equity is now partly a tax asset.
The paper finds that a lender without deposits holds about twice a bank's equity. Affirm holds more than that. It held more than that before the tax release added $1.45bn.
The structure underneath is simple to describe. Secured, fast-turning debt does the funding, convertible notes sit on top, and sold loans spread the rest across outside buyers who can leave after one to three years.
I don't know how much of the 2026 equity a lender or a rating analyst would count. The tax asset is real under the accounting rules, yet it fades in exactly the downturn when capital is needed most. That tension still bothers me more than any other line in this 10-K.