SoFi Technologies (NASDAQ: SOFI)

The call

We are initiating on SoFi Technologies with a HOLD and a $15.00 fair value.

This is the most genuinely two-sided report we have written, and we want to say so at the top rather than pretend the answer was obvious.

The growth is not a story. Adjusted net revenue grew 40.5% year over year, adjusted EBITDA 43.6%, members 35% to a record 15.8 million with a record 1.1 million added in the quarter. That is the nineteenth consecutive quarter clearing the Rule of 40, most recently at 70. Very few businesses of this size are growing like this, and none of it is an accounting artifact: adjusted net income sits only 2.2% above the GAAP figure, which is unusually clean for a company that reports adjusted EBITDA at all.

The problem is what it earns on the equity it holds, and what the price pays for that.

Two businesses, one share price

SoFi is a chartered bank wrapped in a fee-based fintech, so no single framework is honest on its own.

The lending half is a bank. It funds loans with 45.5 billion dollars of deposits, it holds capital against them, and its economics are the spread between what it lends at and what it pays, less credit costs. Net interest income was $788.2 million in the quarter, up 52%, at a net interest margin of 5.98%. For that half, the right anchor is price to tangible book against return on tangible common equity, the same framework we used on JPMorgan.

The other half is fee revenue, $472 million in the quarter, or 39.2% of adjusted net revenue. That is the technology platform and the financial services products, and it barely consumes tangible equity at all. Valuing it on price to book understates it, and we want to be explicit that the bank framework below is therefore harsh on roughly two fifths of the business.

We have still used it as the primary anchor, for one reason: the tangible equity is real, it is there to absorb credit losses, and the returns on it are the thing that has to improve for the current price to make sense. A framework that flatters the company by ignoring its balance sheet would not be much of a framework.

What it earns on the equity it holds

The company guides to $825 million of adjusted net income for 2026 against $9.5 billion of tangible common equity. That is a return on tangible common equity of 8.7%. On second-quarter GAAP net income annualized, it is 6.6%.

We use a 11% cost of equity. That is a judgment and it is the single most important number in this report, so it is stated here rather than buried. It is deliberately generous: a deposit-funded bank with a national charter is less risky than SoFi's 2.19 beta implies, and we have given the company credit for that. A cost of equity derived from that beta would be 14.9%, and every conclusion below would be considerably harsher.

Either way, the spread is negative. SoFi currently earns 2.3 points less than shareholders require on the equity it holds. That is not a criticism of the business. It is what a company looks like while it is still building scale, and the direction of travel is right. It does mean that every dollar of the 2.28 times tangible book being paid today is paid for returns that have not arrived.

Where the book value came from

The bull case's strongest single number is that tangible book value per share rose 55.5% year over year, from $4.72 to $7.34. It is a genuinely impressive figure and it is worth understanding properly.

This matters for how you read the trajectory. A bank that grows book value by retaining profits is self-funding. A bank that grows it substantially by issuing equity is buying growth with dilution, which works while the stock trades above book and stops working when it does not. At 2.28 times tangible book that option is still open. At one times it would not be.

What the price assumes

Solving the identity backwards at a 11% cost of equity and 6% long-run growth, today's 2.28 times tangible book implies a sustainable return on tangible common equity of 17.4%.

That is 2.0 times what the company guides to this year. Not an improvement. A doubling.

Read the top row as the sanity check on the framework: on what it earns today, the bank arithmetic alone values SoFi at $3.94. We do not think that is the right answer, because it ignores 39% of revenue that needs no equity and it ignores a growth rate few companies have. But it is where a pure bank lens lands, and a reader deserves to see it.

The question is how long the gap takes to close.

Two points of improvement a year is a good outcome for any bank, and on that path the price is paying today for a level of profitability reached five years from now, with no allowance for a credit cycle arriving in between.

Our $15.00 sits between the 15% and 17.4% cases. It credits SoFi with reaching returns most digital banks never reach, while declining to pay in full today for returns that require another five years of flawless execution.

The bull case, and why this is a HOLD

We came close to a lower rating and talked ourselves out of it. Here is why.

The de-rating has already happened. The stock is 48.9% below its 52-week high and closer to the low than the high. Whatever the market believed a year ago, it does not believe now. A bearish call is far less interesting after the price has halved than before it.

The fee business is genuinely different. 39.2% of adjusted net revenue needs almost no tangible equity, which means the blended return on capital employed is materially better than the ROTCE figure suggests. As that mix rises, the bank framework becomes progressively less appropriate.

The disclosure is clean. Adjusted net income exceeds GAAP by 2.2%. After writing at length about companies whose adjusted figures flatter a quarter, it is only fair to note when a company's do not.

And the operating momentum is unusual. Record originations of $14.8 billion, deposits up $5.3 billion in a single quarter, cross-buy at 51%. These are not the numbers of a business losing its way.

Why not a SELL. The gap to our fair value is 10.3%. We rated Nike a HOLD at a wider discount. Issuing a SELL here would apply a lower bar than one already on our own scoreboard, and a board whose ratings mean different things at different times is not a scoreboard at all. That constraint is written into the model as an assertion, so it fails the build rather than relying on our restraint.

Why not a BUY. Every route to today's price requires roughly a doubling of returns on equity, and the price pays for it now rather than on arrival.

Risks, and what would change our mind

To the upside. ROTCE improving faster than two points a year. The fee share of revenue continuing to rise, which structurally raises returns because it needs no capital. Deposit costs falling further against the funding they replaced. Any evidence that book value growth is becoming self-funded rather than issuance-funded.

To the downside. A credit cycle. This is a consumer lender whose loan book has never been tested through a genuine downturn at its current scale, and the single number we would watch is net charge-offs against reserves. Continued heavy share issuance, which would mean the growth still needs outside money. Any slowdown in member growth, since the entire thesis rests on the cross-buy flywheel continuing.

To our own reasoning. The cost of equity is doing more work here than in any report we have published. We use 11%. A beta-derived figure would be 14.9%, at which the price implies 26.4% sustainable ROTCE and fair value would be far below ours. We chose the generous end deliberately, and if you think a consumer lender with this loan book deserves the beta-implied number, this report is not bearish enough.

The second soft spot is the framework itself. Applying a bank multiple to a company where 39% of revenue is fee-based overstates the capital intensity of the whole. A sum-of-the-parts treating the fee business on a revenue multiple lands higher than our figure, and a reader who prefers that approach should adjust upward.

Related research

Reports that make a similar argument about a different business. Every call on this site is dated, public and scored the same way, whether it worked or not.

• JPMorgan Chase, the same tangible book framework applied to a bank at the opposite end of its life.
• Palantir, another high-growth name where the question is what the price already pays for.
• Uber, a platform we did back, which is the test of whether growth alone decides our answer.

The full list, with the price each call was made at, is on the scoreboard.

Methodology, sources & disclosures

Methodology

Every figure is computed by a model script rather than typed, and the script runs three self-checks: that the reverse solve reconciles to the observed multiple, that a bank earning exactly its cost of equity values at one times tangible book, and that this rating does not apply a lower bar than a hold already published on our scoreboard. The primary framework is price to tangible book against return on tangible common equity, the same one used on JPMorgan. We have noted in the body that this framework is harsh on the 39 percent of revenue that is fee-based and needs no equity, and that a sum-of-the-parts lands higher.

The reference price

The $16.73 reference price is the close on Thursday, September 17, 2026, verified before any modeling against four independent sources: Morningstar, explicitly timestamped, and Yahoo Finance, CNBC and Motley Fool, each reporting it as the previous close on September 18. The September 18 close of $16.96 was confirmed by two sources but not used, because our convention is the closing price on the trading day before publication and the 17th verifies more strongly.

Primary sources

Second quarter 2026 results, guidance, segment revenue and the tangible book figures from the Form 8-K and earnings release filed July 29, 2026, read directly from the filing rather than through an aggregator. Guidance and the tangible book per share series were confirmed against the earnings call transcript. The decomposition of the book build uses the company's own disclosure that total tangible book rose about 80 percent while per-share book rose 55.5 percent; the difference is issuance, and we have shown the arithmetic rather than asserting it. Company filings at investors.sofi.com; filings as submitted via SEC EDGAR, CIK 0001818874.

Conflicts

The author holds no position in SOFI and has no plans to initiate one, and received no compensation from any party in connection with this report. 2 Comma Investor accepts no payment for coverage, has no banking, advisory or consulting relationships, and does not accept sponsored research.

Scoring

This call is dated and public. It joins the scoreboard as an open position at $16.73 and will be scored against the S&P 500 on September 18, 2027, on the same terms as every other call we have made, whether it works or not. Our standing view on hit rates applies to us before it applies to anyone else.