SoFi Technologies, Inc. SOFI

Key data

The business

SoFi is a bank without branches. It started lending to graduates refinancing student debt, then won a bank charter in 2022, which let it fund loans with its own customers' deposits instead of borrowing wholesale. Three parts. Lending, meaning personal loans, student loans and mortgages, still makes most of the money. Financial Services is the checking, savings, debit card and brokerage. A Technology Platform, Galileo and Technisys, runs the plumbing behind other companies' fintech apps. What a member actually holds is the app, where the paycheck lands two days early, savings pays a high rate, and a personal loan is a few taps away. The moat is the deposit base and the stickiness of a primary checking account. Once the paycheck routes there the customer rarely leaves, and cheap deposits are what let SoFi underprice a stand-alone lender. That is the whole engine: gather deposits at a low rate, lend at a higher one, sell more products to the same member.

The numbers

The story is a business that crossed from burning cash to compounding profit, and the rate of change is still rising.

Revenue has climbed every quarter for two years, from $1.13B to $1.57B, and the last print grew 39% on a year earlier. Profit is the subtler read: net income of $156.6M in Q2 was up 61% from a year ago but down about 6% from Q1, because SoFi is spending harder on marketing and setting aside more against a fast-growing loan book. The turn came in Q1 2024, when full-year losses flipped to profit, and eight straight profitable quarters have followed.

Across the years the shape is unmistakable: revenue compounded about 39% a year from 2022, and the company went from a $360.8M loss that year to a $481.3M profit in 2025. But read the per-share line, not the total. Diluted shares grew about 39% over those three years and reached roughly 1.35 billion in Q2 2026, up about 14% from a year earlier, faster than the pace flagged a couple of weeks back, so the dilution cost of this growth is running hotter, not cooler. Note too that the 2024 profit was flattered by a one-time tax benefit of about $265M, so 2025's cleaner $481.3M, earned against a $44.5M tax bill, is the truer run rate.

For a lender the real revenue is the spread between what the loans earn and what the deposits cost, and that spread is widening.

Net interest income grew 52% in a year, to $788M, faster than the loan book, because deposits are repricing cheaper than wholesale funding as rates ease. That is the lens in one line: the take is expanding, not just the volume. What this memo believes that the tape may not is that the durable earnings power sits in that deposit-funded spread plus the fee segments, not in the loan-sale gains that flatter the headline, and the single print that settles it is the net charge-off rate holding as the personal-loan book scales.

Management

Anthony J. Noto has run SoFi since 2018 and bought about $500K of stock on the open market across May and June 2026, two discretionary purchases rather than a scheduled plan, which is the signal you want from a CEO at $18. Against that, insiders were net sellers of about $2.0M over the year, the largest a $1.82M sale by an officer. Noto's 2024 pay was about $28.1M, almost all stock, roughly 5.6% of that year's profit, rich but tied to a share price that has to work. The weak spot is capital return: the $65.0M bought back in 2025 at an average of $19.94 is a rounding error against the shares handed to employees. Management has hit or edged past its own quarterly guide eight times running, so the guidance record is clean.

How it fails or surprises you

Credit turns on an unseasoned book. The profit engine is personal loans, most of that book written in the last two years and never stress-tested by a downturn. Young consumer books always look pristine until the first real recession prices them. If the net charge-off rate drifts from the high-3s toward 5% as unemployment rises, provisions eat the inflection fast. This is the number the memo can see least well and the one that would prove the read wrong.

Dilution outruns the compounding (downside). Revenue up 39% is real, but shares grew about 14% in the last year and 39% over three. Per-share value leaks steadily until buybacks exceed stock comp, which they do not, not close. Watch the diluted share count each quarter against revenue growth, and the first quarter buybacks finally top issuance.

The platform re-rates it (right tail). If deposits keep flooding in cheaply and the fee arms, Financial Services and Galileo, scale to real contribution profit, SoFi stops being priced as a lender and earns a capital-light platform multiple. The market pays a lender's price today. The tell would show first in segment contribution profit turning sharply positive while deposits compound.

Closing thoughts

One number settles this: the net charge-off rate. If it stays in the high-3s as the personal-loan book scales, the compounding is real and the deposit-funded model compounds earnings faster than the market expects. If charge-offs drift past 5%, the profit inflection reverses and provisions eat the upside. An ambiguous print, charge-offs creeping up while the book still grows, leaves you holding a fast grower with a thin margin of safety. The fatter tail is up because a deposit-funded bank growing members and spread at this rate is worth more than 38 times if the credit behaves, but the downside tail is fatal if a recession prices a young unseasoned book. That is judgment, not a forecast.

The bet is still that SoFi keeps turning a phone app into a real bank, taking in customers' deposits cheaply and lending that money back out at a profit. What breaks it is charge-offs rising faster than the loans grow, or the share count outrunning revenue. Watch the net charge-off rate against loan growth, and diluted shares against revenue growth each quarter. If charge-offs cross 5% while deposits stall, the story is wrong, and those two numbers will tell you before the price does.