SoFi Technologies closed the second quarter with 388,336 cumulative crypto products on its books. The same filing reported $1.183 million of net crypto transaction revenue. Read those two numbers together and you have one of the most honest — and most damning — accounting disclosures of the year.
The gross revenue line was $134 million. Cost of crypto transaction revenue was $133 million. The difference, $1.183 million, is 0.88% of the gross line. That is not a profit margin. That is a rounding error with a bank charter attached.
We didn't need another earnings press release to know where this was heading. We needed the cost line. SoFi gave it to us, almost accidentally, buried under the pageantry of a consumer fintech earnings package.
SoFi announced the phased launch of consumer crypto trading on Nov. 11, 2025. Q2 is therefore one of the first full quarters of that product's existence. These numbers are not trailing noise from a legacy business. They are the opening statement of a new one — and the opening statement is thin.
The accounting treatment matters more than the headline. In its first-quarter Form 10-Q, SoFi said it records crypto transactions on a gross basis because it acts as principal. It buys digital assets from, or sells them to, third-party liquidity providers before transferring the assets to or from member accounts. SoFi books money from member buys and sells in its gross crypto revenue line, along with transaction fees after rewards. Most of that money flows straight back out to cover the assets SoFi buys for members and the payments tied to member sales. What remains is net crypto transaction revenue, driven mainly by the fees SoFi collects for handling each order.
This is the distinction between a principal and an agent, and it is the entire story of retail crypto economics in one paragraph. The gross line is a volume meter. The net line is an economic meter. SoFi is telling us, with remarkable precision, that the volume is enormous and the economics are not.
The net line improved sequentially from a small base. Q1 produced $852,000 of net crypto transaction revenue on $121 million of gross revenue, offset by $120 million of transaction costs. Q2 net rose to $1.183 million — an increase of roughly 38.8%. The first-half total reached $2 million.
One caution, and it is a governance caution: the two numbers relate to different time periods. The 388,336 tally covers every crypto account opened through quarter-end. The $1.183 million covers Q2 revenue alone. Mixing the two to compute a per-user take rate would be malpractice. SoFi's own disclosure structure prevents that calculation. That may be precisely the point.
This disclosure structure is not accidental. SoFi answers the question that makes crypto look like growth — how many products — and avoids the question that would expose it as a cost center — how much net value. The asymmetry is the tell. Every line of the filing is chosen, and the absent lines are the ones that matter.
Every line of code writes a history of power. So does every line of an earnings filing. SoFi's filing writes a specific history: a consumer financial giant that wants to hold 388,336 crypto accounts without building a crypto profit center.
Start with the accounting. Gross-basis reporting hides the actual economics. When SoFi books $134 million of gross crypto revenue, it is counting the full dollar value of every trade its customers executed. If a member buys $100 of Bitcoin, SoFi records $100 of revenue and roughly $100 of cost. The gross line measures how much retail money moves through the pipe, but it says nothing about how much value SoFi actually extracts. The net line is the extraction. $1.183 million on $134 million of flow is a fee yield of 0.88%.
Based on my experience auditing early Ethereum ICO contracts in 2017, I can tell you the discipline is identical in principle: follow the token flow until the value accrued to the firm becomes visible. In those contracts, the reentrancy vulnerabilities were hiding where the money moved twice. Here, the vulnerability is not in code — it is in the cost line. $133 million of cost means SoFi is passing almost all of the spread back to wholesale liquidity providers and to member rewards. The retained economics is the order fee. That makes SoFi a payment processor, not a crypto business. Valuation multiples should follow accordingly.
There is also a regulatory dimension that makes the thin net line more expensive than it appears. Operating a crypto trading desk under a bank charter triggers capital treatment questions, consumer protection obligations, and examinations that a pure fintech would not face. SoFi is running two compliance regimes for one product. The net revenue line does not include that overhead, which is precisely why a standalone profit figure would be the only honest number.
The 38.8% sequential growth deserves scrutiny, not applause. Going from $852,000 to $1.183 million in net revenue is improvement from a near-zero base. But consider the build cost. A consumer crypto brokerage requires custody integration, AML and KYC infrastructure, tax reporting, a liquidity provider framework, and permanent legal exposure. None of that is cheap. SoFi does not disclose a standalone crypto profit figure. For a company that discloses its net revenue with such precision, the absence of a profit figure for this product line is a deliberate choice. Truth emerges from transparency, not from silence. The silence here is a ledger entry of its own — an acknowledgment that the product, on a fully loaded basis, has not earned its place on the income statement.
Then there is the account count. 388,336 cumulative crypto products tells you how many accounts were opened, not how many are active, not how many hold balances, not how many traded in Q2. In my work designing Aave's governance framework in 2020, we learned that headline participation metrics lie. We adopted quadratic voting not because it made adoption numbers look big, but because it made power distribution legible. SoFi's disclosure does the opposite. It optimizes the legibility of growth and obscures the distribution of revenue. Without an active-user metric or an average fee per trade, the 388,336 figure is a signup counter, not a business metric.
The pricing gap sits in the principal model. Because SoFi acts as principal, it absorbs inventory risk between the wholesale liquidity market and the member order. If Bitcoin sells off and member sell orders outpace the desk's hedging, SoFi eats the slippage. The $133 million cost line will not stay flat in a crash. It will spike. The 0.88% net yield will compress to zero or negative. That is the Terra-Luna lesson applied to brokerage accounting. In 2022, I watched institutions confuse gross volume with resilience. The protocols that survived were the ones that stress-tested their spread under extreme flow. SoFi's filings contain no stress scenario, no sensitivity table, no disclosure of inventory positions. The entire model banks on orderly markets.
None of this matters, however, until you reframe what the business actually is. $2 million of half-year net revenue is not a crypto business. It is a customer acquisition cost that happens to occupy a revenue line. Think about what $2 million buys in this industry: one senior engineer's annual compensation package. A fraction of a single marketing campaign. SoFi is not building crypto to earn from crypto. It is building crypto to keep 388,336 accounts inside its deposit, lending, and payments ecosystem. The product is the hook. The bank is the margin.
That reframing changes every conclusion you would draw from these numbers. It also connects to a broader pattern: major financial institutions are building the rails, the custody, and the infrastructure to profit from roughly 13.9 million Bitcoin they do not own. They do not need the public chain to be their profit center. They need the chain to be their acquisition funnel.
The contrast with Robinhood is instructive. Competitors have reported crypto revenue dropping by hundreds of millions as retail rotates out of trading. SoFi's gross line keeps flowing because its members are not traders; they are depositors who occasionally hold assets inside the app. The account relationship is the moat, not the execution engine.
This is where my skepticism about the institutional narrative firms up. For three years, the market has told a story about traditional institutions embracing on-chain finance as an extension of decentralization. The SoFi filing tells a blunter story. Institutions will offer access to crypto when it is gross-revenue cheap and net-revenue sticky. They will not subsidize the chain. They will extract the customer. Governance isn't a dashboard of product counts; it is the audit trail of who gets paid last. In SoFi's disclosed structure, the answer is clear: the customer gets the product, the liquidity provider gets the spread, and SoFi gets the deposit.
The counterintuitive conclusion is that these disappointing numbers are among the healthiest signals in the sector. Everyone wants the headline — 388,336 products — and nobody wants the reconciliation. But the reconciliation is the truth. A 0.88% net yield forces honesty about what retail crypto actually monetizes. It also exposes the rest of the industry that refuses to file the same reconciliation.
Think about the Layer2 ecosystem. We now have dozens of rollups serving the same small user base, each reporting gross activity, each avoiding the net-value question. That is not scaling. That is slicing already-scarce liquidity into fragments and calling fragmentation progress. SoFi's filing is the same phenomenon in a bank costume: gross revenue inflated by principal accounting, net revenue thin as tissue, and a product count standing in for actual usage.
The deeper lesson is for analysts and token holders. If a listed fintech with 388,336 crypto accounts cannot convert that base into meaningful net revenue, then every protocol reporting gross volume as a proxy for value is performing the same accounting trick with less disclosure discipline.
The reflexive take is that SoFi's crypto line is failing and retail is abandoning digital assets. The more accurate take is that retail crypto trading was never the product. The product is the account relationship. SoFi is not trying to win the trading race that leaves competitors reporting collapsing crypto revenue lines. SoFi is trying to win the deposit race, and the crypto product is the cost of admission.
We should stop asking whether SoFi's crypto line will become profitable in the way analysts frame profitability. It will not. It will become a deposit machine with a thin fee overlay. That is not failure. It is a pivot disguised as a product — and the market has not yet learned to read it.
The filing's real content is the structure it reveals. 388,336 accounts. $2 million in half-year net revenue. Zero standalone profit disclosure. SoFi will keep growing the account line while the net line stays thin, because that is the design. The question for every other institution building crypto rails is whether they have the discipline to show the spread, or the incentive to bury it. Every line of code writes a history of power. SoFi just wrote this quarter's entry. It reads like an acquisition funnel, not a profit center. The market should price it accordingly.