Flagstar Financial, Inc. FLG

Key data

The business

Flagstar Financial, until last year New York Community Bancorp, is a roughly $88B regional bank whose loan book was built on rent-regulated New York City apartment buildings. That concentration was the whole company for decades and nearly ended it in January 2024, when a surprise multifamily reserve and a dividend cut triggered a deposit run and a collapse. The bank sold its mortgage-servicing and warehouse operations to Mr. Cooper, took a $1.05B capital injection, and changed its name and its management. What is left is a slimmed-down commercial bank, still roughly $61B in loans, deliberately shrinking its apartment and commercial-property exposure while building safer business and private-bank lending. The customer is a New York landlord or a mid-market business owner. The deposit base, about $67B, is the franchise that has to hold while the loan book is rebuilt.

The numbers

The shape here is a bank pulling out of a crater. Five straight quarterly losses through the third quarter of 2025 gave way to three consecutive profits, small but real. The 2023 revenue line is distorted by one-time gains from the Signature Bank deal, so ignore it. What matters is the recent run: net revenue steady around $500M a quarter while the loss line flipped positive.

The inflection is the fourth quarter of 2025, the first profit after the bleed, and it has held for three quarters. Q2's $34M topped Q1's $21M, and net interest income has stayed near $440M even as the loan book shrank, which means the margin is widening as costly wholesale funding rolls off. The beats and misses against forecasts have been mixed and small, a reminder the recovery is not yet linear.

Put the arithmetic plainly. The bank earns on roughly $7.8B of tangible net worth, about $18.70 a share. Q2's $34M annualizes to $136M, a 1.7% return on that net worth, a rate that would keep the stock cheap forever. The market is not paying for 1.7%. Consensus puts 2027 earnings at $1.33 a share, near $555M, about a 7% return, and prices the stock at 10 times that number and 0.7 times net worth. So you are handed a wager the machine has already framed: profit must roughly quadruple from here as loan losses fade and the balance sheet stops shrinking. What this memo believes that the tape does not is narrow. The early-delinquency collapse, multifamily arrears down 60% in six months, says the 2027 snap-back is likelier than a 0.7x multiple implies, and the quarterly provision line is the print that settles it.

The credit arc is the whole story, shown here on the bank charter (FDIC Call Report).

Net charge-offs have more than halved from the late-2024 peak, and noncurrent loans, having climbed to 5.5% by late 2025, are now rolling over, 4.78% on the charter and 4.59% at the holding company by June, down from 4.90%. The tell that the worst is passing sits in the early bucket: multifamily loans 30 to 89 days past due fell from $588M to $233M in six months. Reserves are easing in step, which is either vindication or complacency, and that is the second linchpin.

Management

Joseph Otting, the former U.S. comptroller of the currency who ran OneWest, took the top job after Steven Mnuchin's Liberty Strategic led the $1.05B rescue in March 2024. The record since is a demolition crew: multifamily and commercial real estate loans cut hard, wholesale borrowings down from $13.2B to $6.7B in a year, deposits steadied near $67B after the run. That is the right work, and it shrinks the bank on the way through. Insiders bought none of their own stock in the last twelve months, and none was sold either, so the recovery still rests on the 2024 raise, not on fresh open-market conviction. Compensation detail is not in this filing set. There are no buybacks, and at 0.7 times book there should be, once capital allows.

How it fails or surprises you

The rent freeze bites (downside). The 10-Q blames part of its higher charge-offs on New York City's move to freeze rents on multifamily buildings. Roughly $61B of loans still lean on rent-regulated apartments whose cash flows just got capped. If freezes spread or deepen, the reserve relief that just began reverses and the profit streak ends. Watch the multifamily non-accrual line each quarter.

Coverage is thinning as they charge off (downside). Reserves against non-accrual loans fell to 31% from 35% in six months, and reserves to loans slipped to 1.57%. Management reads improving credit, a skeptic reads a cushion being spent. About 40% of non-accrual loans are current on payments today, a number that flatters and can flip fast. One bad multifamily quarter tells you which reading was right.

The re-rate runs ahead of the earnings (right tail). At 0.7 times net worth the market demands proof of normal profit before it pays up. If Otting posts even 8% return on tangible book, roughly $1.50 a share, well before 2027, the stock re-rates toward book on the earnings and again on the multiple. Two clean quarters of falling non-accruals and a rising margin is the tell, and nobody is paying for it now.

Closing thoughts

The quarterly provision for credit losses is the number that decides whether this recovery is real. Each quarter it stays contained, the path to normalized earnings gets clearer and the 0.7x multiple starts to look too cheap. Each quarter it spikes, you learn the apartment-loan bleed has not stopped and the turnaround takes longer. The left tail is a multifamily relapse, made live by the rent freeze, that forces fresh reserve builds and pushes profitability out another year. Capital is adequate, so that tail means dilution and delay, not solvency. The right tail is a clean normalization that re-rates a 0.7x bank toward 1.1x on tangible book of $18.70. The downside is deeper than the last two quarters suggest, but the early-delinquency collapse says the fatter tail now points up.

The bet is still that Flagstar has stopped bleeding on New York apartment loans and can earn a normal profit on its net worth again. What breaks it is the multifamily book, and the pair to watch is dead simple: non-accrual loans as a share of the total, 4.59% and falling, set against the provision the bank books each quarter. While non-accruals keep dropping and the provision stays contained, the turn is real. The quarter both turn back up is the quarter the story was wrong.