Key insights
- The author argues SoFi could resolve the debate around its fair value adjustments by complying with GAAP (ASC 820) and disclosing gains/losses specifically from these adjustments. The lack of transparency around realized vs. unrealized gains raises concerns about the quality of SoFi's earnings and potential overvaluation of assets. This could negatively impact investor confidence.
I have a negative position in SoFi and this is not financial advice.
We've heard a lot of back and forth over the years regarding the validity of SoFi's fair value adjustments. One critical piece that has been overlooked is that SoFi has long had the ability to effectively resolve the debate, through the disclosure of information regarding gains/losses attributed specifically to fair value adjustments. In fact, this disclosure is required by ASC 820-10-50-2(c), a part of Generally Accepted Accounting Principles (GAAP) which public companies must comply with presumptively under SEC regulations.
Under ASC 820-10-50-2(c), a reporting entity is required to provide a rollforward table for recurring Level 3 measurements disclosing separately the "[t]otal gains or losses for the period recognized in earnings." Crucially, to ensure investors can evaluate the nature of those gains, ASC 820-10-50-2(d) strictly requires the disclosure of:
"The amount of the total gains or losses for the period in (c)(1) included in earnings... that is attributable to the change in unrealized gains or losses relating to those assets and liabilities held at the end of the reporting period..."
The purpose of ASC 820-10-50-2(d) is to facilitate investor understanding of earnings "quality," specifically as it relates to Level 3 "modeled" fair value adjustments. Realized gains represent actual, verified cash received when an asset is sold or settled (higher quality). In contrast, unrealized gains are entirely attributable to management's internal valuation models, representing purely paper-based markups on assets still held on the balance sheet (lower quality).
If a company persistently reports positive unrealized gains, but negative realized gains, that combination would strongly suggest that the company’s fair value models are persistently overstating the fair value of assets, with the losses on persistent overvaluations “bleeding back” as realized losses, as the overvalued assets are sold or amortized.
Without the explicit delineation required by the rule, investors cannot track the historical pattern of positive or negative discrepancies between these realized and unrealized marks. This historical discrepancy is the critical metric used to determine if prior unrealized gains were artificially inflated by the model, or if the model accurately reflects economic reality.
Right now, SoFi does not disclose the breakdown between unrealized gains and realized gains. That means investors do not have an understanding as to whether the model's track record has been correct.
This disclosure rule is not ambiguous or complicated. Deloitte (SoFi's auditor) has a publication called Deloitte Accounting Research Tool has an illustrative example showing how to comply with the rule:
In this you can see that $7 of the gains were unrealized - in the case of SoFi that would mean such gains were model based.
SoFi's rollforward table has an "impact on earnings" column is essentially an aggregation of all the gains or losses attributable from changes in the other columns. There is no detail provided regarding what gains were realized or not realized.
SoFi has a note under the table that basically explains the "Impact on Earnings" column is an aggregation:
This GAAP issue has been conveyed by me to SoFi multiple times over the last month, without correction. Under SEC rules, once a company is aware of a material deficiency in its reporting, it needs to essentially cure it within 4 business days (i.e. through a corrective amendment). The lack of a cure could only be legally justified if SoFi thinks its disclosure follows GAAP. Assuming SoFi thinks its presentation is GAAP compliant, it will be in serious trouble with the SEC the longer it fails to correct the issue.
I personally find the lack of correction very concerning: the rule is fairly black and white: this is not one of those accounting rules where there is a lot of grey area or judgment calls. It is a simple mathematical disclosure. If you go by Deloitte's own handbook, compliance seems straightforward.
Importantly, making the disclosure does not require any judgment calls about the validity of the model - it just would show investors whether the model's "predictions" on fair value have been validated as gains are realized with loan sales or loans amortizing
The fact that Sofi is audited does not somehow absolve this mistake. Mistakes get made in audited financial statements. An audit opinion is not an absolute guarantee of perfection. There is a well established protocol and rules dealing with corrections after the fact, which are fairly routine. The crucial thing is that everyone has to act quickly once the error has been flagged.
In all events, investors should want SoFi to comply with ASC 820-10-50-2(c),
- It is legally required. 2. If realized gains are flat or positive, it would show that SoFi's models are accurate, because SoFi has not been experiencing losses that were not baked into the model. 3. This would give investors peace of mind if they have even a little bit of concern about the Muddy Waters report.
One other bit of information that is relevant regarding fair value adjustments that is not in SoFi's SEC filings is the total amount of fair value adjustments that actually flow into earnings. That number is impossible to find, and impossible to extrapolate accurately, with the information provided just in SEC filings
But that information is disclosed in SoFi's Federal Reserve filings. For 2025, fair value adjustments, net of associated hedges, was $538 million.
The associated hedges appear to have been -$164 million in 2025 (this is taken from FY 2025 10-k, page 200).
Based on these two pieces of information, it looks like somewhere between $538 million and $702 million in earnings were attributed to "adjustments" out of total gross revenue of $4.7 billion ($3.4 billion in interest income, $1.4 billion in noninterest income).
Obviously fair value adjustments are important in the grand scheme of things, as that is between 10-15% of total revenue. So that only kind of emphasizes why SoFi should be complying with ASC 820-10-50-2(c) - if realized gains are negative, and unrealized gains are positive, that would mean even though fair value adjustments is a big number, the positive portion of it may not be "real."
I have two final comments regarding some good questions I have seen in the wake of the Muddy Waters report. One is how could SoFi could even be experiencing fair value losses, considering its loan default rate is fairly low, and given that SoFi borrowers are generally super prime.
The question is super reasonable. How could a loan lose value if the borrowers are paying it back? The reason is because "fair value" is defined very technically: it represents the market value of the loan - the amount a third party would spend to purchase the loan.
Fluctuations in market interest rates depress the value of pre-existing loans originated under prior market conditions when interest rates were lower. That makes complete sense: why would someone buy a treasury bond paying 2% interest over 10 years from a third party if the same person could purchase a 4% interest bond over 10 years straight from the government?
The treasury department maintains a daily database of marketplace prices of treasury bonds/notes, so you can see how bonds are doing against par on a daily basis. If the bond/note is trading above par, that means there is a premium associated with it - buyers are paying more than the face value of the loan to purchase it. If less than par, that means, buyers are expecting a discount to purchase the note/bond. Importantly, this is actual marketplace data - not a model.
When you compare actual treasury marketplace data on various notes, you can see that, as expected, they have declined in value when interest rates fluctuated. In contrast, SoFi's personal loans and student loans kept a premium, even though those loans are not immune to the inevitable economic impact of hikes in interest rates decreasing intrinsic value.
The below charts compare market based valuations of treasury notes (2 and 5 years over time) with model based valuations of SoFi's loan inventory (note that some of the 2 year notes amortized during the time period, so their marketplace values do not go to the end).
You can see a similar phenomenon when comparing SoFi's fair value models to those of other banks/lenders. Below chart is through Q3 of 2025.
SoFi's fair values show a persistent premium, when everyone else fluctuated, and went down hard in 2022/2023. Note the comparisons are mostly to mainstream banks like Bank of America, Wells Fargo, PNC, etc.
So regardless of what you think about the debate, the hard data is this: SoFi's models treat SoFi loans are being essentially unaffected by interest rate fluctuations that caused large and marketplace demonstrated changes in the value of risk free deb