PennyMac Financial Services, Inc. (PFSI) stock intrinsic value fair value estimate & owner earnings analysis

PennyMac Financial Services originates residential mortgages, mostly by buying loans from independent lenders (correspondent lending, $111.3 billion in 2025), plus its own call-center refinancing and broker channels, and it services mortgages for a fee. It keeps the servicing rights on much of what it sells, and those rights were attached to loans with a balance of $462.0 billion at the end of 2025, up from about $426 billion a year earlier. It also fulfills loans for and subservices for its affiliate PennyMac Mortgage Investment Trust. How we measured earnings. Reported operating cash flow is meaningless for a mortgage lender: it swings from about $(6.2) billion in 2020 to about +$6.0 billion in 2022 simply because loans are bought and sold. We removed those loan-inventory flows, the non-cash gain booked when servicing rights are created, and other working-capital lines, and kept the run-off in value of the servicing rights and the cost of hedging them as real costs. We subtracted stock compensation and capital spending (mostly capitalized software; all treated as maintenance). The result is earnings per diluted share of $4.21 (2019), $21.07 (2020), $13.79 (2021), $7.23 (2022), $3.41 (2023), $7.05 (2024) and $10.01 (2025). The share count includes exchangeable units. The 2020 and 2021 refinance boom came from near-zero interest rates that have since reversed, so we set them aside. Our starting level is the median of 2019 and 2022 through 2025: $7.05 per share. We did not use 2025's $10.01 as the starting point: half-year 2026 net income was $104 million versus $213 million in the first half of 2025, so we would want more than one good year before moving the level up. Trailing owner earnings grew about 42% in 2025; our base case assumes only 3% in year 1, below that, so no reversal of a slowdown needs to be argued. Base case, about $108 per share. Earnings start at $7.05 and grow 3% a year for five years and 2% for five more. The evidence for that growth: the balance of loans behind our servicing rights kept rising through the 2022-2023 downturn in mortgage volume and grew about 8% in 2025, funded from the company's own production, and the company sold $607.7 million of servicing rights to outside buyers in 2025, showing the asset can be turned into cash. We assume the level then persists for another 15 years (through year 25), reflecting a franchise that has been profitable every year in our seven-year record; after that we credit nothing. We do not credit market share gains, new products, or lower rates. Bear case, about $27. Earnings stay at the 2023 trough of $3.41 for ten years, then fall to half that for five more before ending. This reflects a business that is heavily financed by short-term warehouse borrowings, a servicing asset whose value is marked to model and can fall when prepayments rise, and a 2023 result that shows how low earnings go when refinancing volume dries up. It is built on a result the company has already posted, not a hypothetical collapse. Bull case, about $170. Earnings start from the 2025 level of $10.01 (achieved at today's rate environment) and grow 5% a year for five years, then 3% for five, and persist through year 25. Growth is set below the roughly 8% recent growth of the servicing book. Nothing here relies on a return to a refinance boom. What we tested. A large share of reported profit is a non-cash gain from creating servicing rights ($2.94 billion in 2025), and the company's cash from selling loans was negative once that gain is excluded. The $607.7 million of outside servicing-rights sales in 2025 and the hedging cash flows show the assets are marketable, but the value rests on the model inputs that produced a $251.7 million valuation loss in 2025. That is why the bear case is wide. A dispute with a software provider, Black Knight Servicing Technologies, was settled in prior years and appears to no longer weigh on results. Accounting judgements. Interest-rate hedging results and servicing-rights fair-value changes were kept in earnings rather than added back. Earnings before 2019 were not used because the exchangeable-unit structure makes per-share figures not comparable. We could not reconcile why first-half 2026 profit fell by half while revenue rose 19%. What would make this wrong: a period of sharply higher prepayments that shrinks the servicing book faster than production replaces it, tighter warehouse financing, or a loss of the affiliate relationship that supplies fulfillment and subservicing volume.