Arbor Realty Trust arrived at the second quarter of 2026 in the middle of a commercial real estate credit cycle that has already cut its stock more than half - from a 52-week high of $12.58 to about $5.35 - and turned the company's central bet into the question the whole report answers. The bet was that Arbor's strongest earnings engine, the Structured Business that originates and holds bridge loans on multifamily and single-family rental properties, could keep generating interest income thick enough to cover the mounting credit losses on its legacy book. That engine stalled this quarter. GAAP net loss attributable to common stockholders was $37.3 million against a year-ago $24.0 million of net income - diluted EPS of $(0.20), labeled GAAP, versus $0.12 - and the loss was contained almost entirely inside the Structured segment. Credit provisions more than doubled to $38.2 million, an impairment on real estate owned added $13.7 million, and the segment swung to a $49.0 million loss attributable to common. Meanwhile the Agency Business, which originates, sells and services Fannie Mae and Freddie Mac multifamily loans for a fee, kept earning money - $8.4 million in the quarter and a $36.7 billion servicing book. What the headline says is that the earnings engine broke; what the split says is that the fee engine did not.
The more consequential number is the one the dividend answers to. Distributable earnings per share - the REIT measure Arbor itself says it weighs in setting the dividend - fell to $0.10 in the quarter from $0.25 a year earlier, yet the board declared a $0.17 quarterly dividend. The dividend now runs ahead of what the company is earning on the measure it cares about. That spread is the tension at the center of the whole story: a management that has answered the credit cycle by buying back stock at roughly half of book value - $114.3 million of shares at $5.42, or 49% of book, in July alone, after $20.8 million at $5.85 in the quarter - while paying out a dividend its distributable earnings no longer fully cover. The buyback is the argument that the book is worth more than the market will pay; the dividend is the price of holding on while the market decides. The falsification framework below spells out which number breaks first.