FICO lost more than half its value in two months. Its scores didn't get worse and its revenue kept growing. What changed was the rule that forced lenders selling loans to Fannie Mae and Freddie Mac to buy them. This is how that happened, what could come next, and the seven checks I now run so I don't get caught by the same pattern.
Source: IBKR daily bars. 29 Sep close $617.87, intraday low $595.19.
For years, loans sold to Fannie Mae and Freddie Mac had to carry a Classic FICO score. Lenders had no choice, so FICO could raise its mortgage price again and again. Reports on the latest quarter note that mortgage growth came from price, not from more loans.
If one pricing table treats both scores alike, a lender who picks VantageScore gets the same loan pricing for a fraction of the score cost. The chart below shows what that gap looks like on a single mortgage application.
Traditional model: $10 per score. Performance model: $4.95 per score plus a closing fee FICO describes as $33 "per borrower per score". FICO 10T: $0.99 per score plus a $65 closing fee. This chart counts each closing fee once; if it applies to all three scores, the performance model is $113.85 and 10T is $197.97. VantageScore 4.0: $0.99 per score, and TransUnion includes it at no extra charge through 2026 for mortgage customers who also buy a FICO score. Resellers add their own fees on top. FICO's direct-license prices were not yet live as of July 2026.
These are pulls that include VantageScore, not loans underwritten with it. TransUnion bundles it free with a FICO purchase, so many are likely lenders taking both scores. The direction is still clear.
Scale runs 0x to 60x. "2023–25" is trailing GAAP P/E, which ran above 50x each year and hit 93x in Sep 2024 (FullRatio). "Mar 2026" is the forward multiple cited by Investing.com. "Now" is the $617.87 close against FICO's FY26 guidance of $42.43 non-GAAP EPS; on the $36.86 GAAP guidance it is about 17x. On 29 Jul, before the fall, the same measure was about 32x. When a premium multiple meets a threat to pricing power, both the earnings and the multiple fall together.
Almost everything depends on one line of the income statement. Last quarter, mortgage origination scores brought in $283M, 71% of score sales to lenders and about 42% of all revenue. They cost almost nothing to produce, so they carry most of the profit.
My own estimate, not company guidance. It assumes mortgage scores earn about a 90% margin, and applies the company's 62% non-GAAP operating margin, $5.6B of debt at 5.64% interest, and a ~24% tax rate, and charges all interest expense to the rest of the business. Sharing interest in proportion to operating profit instead would give roughly $30 from mortgage scores and $19 from the rest. The rest covers auto, card and consumer scores plus software.
My read. Weighting the three cases 25 / 50 / 25 (EPS × multiple in each) gives about $577, a little below its $617.87 close on 29 Sep. The market isn't obviously wrong, and there is no margin of safety yet. FICO also shows red on five of the six company checks in the stress test below, so I'm watching, not buying, until the November guidance.
Scorecard. I don't own FICO. After the FY Q4 report in early November I'll come back to this page and score these scenarios against what actually happened.
A moat that comes from a rule can be taken away by a rule. Before paying up, ask: if the rule disappeared tomorrow, would customers still choose this product?
FICO passes in credit cards and auto loans, where lenders choose it freely. In mortgages, where a rule did the choosing, that is still unproven.
Check how much growth comes from raising prices versus selling more. If those price increases land on a bill voters already complain about, expect someone in government to notice.
Mortgage score prices rose sharply during a national argument about housing affordability. FHFA Director Bill Pulte then criticised FICO's pricing in public, saying FICO "had taken their prices from $4.95 up to $10".
If the companies that sell your product also own a competing one, they can switch their effort the moment the rules allow it.
The three credit bureaus sell FICO scores and co-own VantageScore. TransUnion added VantageScore free for mortgage customers buying a FICO score, and the share of its mortgage pulls that include VantageScore went from under 5% at the start of 2026 to about 30% in Q2.
At 40–50x earnings, a stock is priced for years of unchallenged pricing power. When that is questioned, earnings estimates and the multiple drop at the same time.
Trailing P/E was above 50x in 2023–25. It is now about 15x on this year's guidance, while revenue was still growing 26%.
Historical multiples describe the old business. Once the business model changes, "lowest in ten years" says little about where the bottom is.
In March 2026, about 24x forward earnings was called a 10-year low, with an average target near $2,068 and the stock around $1,100. It has since fallen about another 44%.
A big one-day drop caused by the market is noise. A big drop caused by news that breaks the thesis is information. Reduce or exit, and buy back only on new evidence.
The 4 Sep and 29 Sep drops were each the regulator taking another step in the same direction. Anyone averaging down after one step took a bigger loss on the next.
A company that borrows to buy back its own stock at a premium price is betting that nothing changes. If the business then weakens, the debt stays while the shares it bought are worth half.
In the June quarter FICO spent $1.96B buying its own stock at an average $1,149, partly with a new $1.5B loan. Debt is now $5.6B, net debt is 3.7x EBITDA, and the stock is at $618.
Run this before adding any "wide moat" name, and again when a position gaps down 15% or more on company news.
| # | Question | FICO mortgage, as of 29 Jul |
|---|---|---|
| 01 | Would customers still choose it if no rule required it? | Unproven |
| 02 | Is most recent growth from price rather than volume, on a politically visible bill? | Yes |
| 03 | Does a distributor or partner own a competing product? | Yes |
| 04 | Is the multiple well above the market (35x+)? | Borderline, 32–37x |
| 05 | Has a regulator or politician named the company's pricing in public? | Yes |
| 06 | Is there a written exit rule for a thesis-breaking gap? | Investor check |
| 07 | Is management borrowing to buy back stock at a premium multiple? | Yes |
Row 04 uses FY26 guidance at the 29 Jul close: about 32x non-GAAP, 37x GAAP. Row 06 is about you, not the company. Two or more red answers: size smaller, write the exit rule before buying, and track the regulator's calendar as closely as the earnings calendar.
The moat is mostly the network itself, which passes check 01. But card fees are a political target, so checks 02 and 05 deserve a look before paying up for them.
Mostly passes
Rising electricity bills tied to data-center demand are becoming a political issue, and grid operators have capped capacity-auction prices before. That is check 02: growth from price on a bill voters see.
Watch closely
The best moats survive a rule change. Own the ones customers would choose anyway.
29 Sep 2026, after first publication: close updated to IBKR's final bar ($617.87, −26.5%); Direct License Program described as announced, not live; the $33 fee quoted in FICO's own words; TransUnion's free offer limited to FICO buyers; the VantageScore figure described as pulls that include it; leverage described as net debt to EBITDA; the ~20-point finding and the MBS point attributed to TD Cowen; the 30 Jul drop no longer attributed to the regulator; stress-test rows 01, 04 and 06 corrected; multiples given with their basis and sources. The scenarios, odds and price ranges are unchanged.