PFSI

PFSI screens cheap on the headline: $83.35 buys you a mortgage banker earning $9.30 of diluted EPS (P/E ~9.0), with net income up 60.9% YoY to $501M and book equity of $4.31B against a $4.35B market cap — i.e., the stock trades at roughly 1.0x book. ROE of 11.6% is respectable for a rate-sensitive mortgage franchise, and the 10-K shows the earnings rebound was largely driven by $551M of Level 3 pre-tax fair-value gains in 2025, much of it the +$453.8M interest-rate-lock-commitment mark and a +$362.4M swing on mortgage servicing liabilities. That is real GAAP income, but it is mark-driven income, not cash — and the contrast is stark: operating cash flow was -$1.65B in FY2025. The headline P/S of 216x is a definitional artifact (the disclosed contract-revenue line excludes the loan/MSR gains that actually drive the P&L), so it should be ignored, not feared.

The MD&A is the reason this can't be a clean buy. Management itself flags that $10.1B (34% of total assets) sit in Level 3 with significant unobservable inputs, and a further $9.6B in Level 2 — together 68% of assets carried at fair value. The 10-K is explicit that 'different persons in possession of the same facts may reasonably arrive at different conclusions' on these marks, and that subsequent transactions may clear at values 'significantly different from those reported.' Layered on top: $6.23B of long-term debt (up 18.6% YoY) plus $2.45B of short-term borrowings, liabilities/equity of 5.82x, and a Risk Factors section that leads with interest-rate sensitivity, MSR prepayment/delinquency risk, covenant risk on secured financing, and dependence on the MBS market remaining liquid. The +$362M swing on MSR liabilities is the kind of line that can reverse if rates fall and prepayments accelerate.