DYNEX CAPITAL INC (DX) stock intrinsic value fair value estimate & owner earnings analysis

Dynex Capital owns a portfolio of mortgage securities guaranteed by Fannie Mae, Freddie Mac and Ginnie Mae, about $26 billion of assets at June 30, 2026, financed mostly with about $22.6 billion of short-term repurchase borrowings and hedged with interest-rate swaps and Treasury futures. There is little credit risk in the securities; the risks are interest rates, the spread between mortgage yields and hedges, and the leverage of roughly seven to eight dollars of borrowing per dollar of equity. A share is worth what the portfolio is worth per share and the return the company can keep earning on it, so we value it on book value and on its demonstrated return on that book rather than on operating cash flow. Cash flow from operations measures only the interest carry and misses every gain and loss on the securities and hedges, which is most of the story; on that measure the per-share figure fell from about $12.50 in 2016 to under $1 in 2023 through 2025 even as the share count grew more than tenfold. The record. Over the ten years to 2025, shareholders received $18.79 a share in dividends while book value per share fell from $23.14 to $13.45, a true return of about 5.3% a year. 2025 was a strong year: book rose $0.75 and the $2.00 dividend was paid, a 21.6% return on starting book, helped by spreads tightening after the company bought heavily at wide spreads. In the first half of 2026 book slipped again to about $12.90 (equity of $3.18 billion less $111.5 million of preferred stock, over 237 million shares), so the 2025 gain has partly reversed. Since 2023 the company has more than quadrupled its share count by selling stock through an at-the-market program; this has been done near book value and has not diluted book, but it means growth in size is not growth per share. Base case ($9.68): starting from book of $12.90, the company earns about 4.5% a year on book over time, below the decade's 5.3% because that average leans on the exceptional 2025, and roughly in line with what the business returned in the ten years excluding it. No growth in book per share is credited, because the company has not produced any over a decade. Bull case ($12.89): the wide spreads at which capital was deployed in 2025 and lower short-term funding costs persist for two years, producing returns near 10% on book, then settle to 5.5%; even this is worth about book value and no more, because nothing in the record shows the company compounding book. Bear case ($5.45): one full shock of the kind Agency mortgage REITs have actually had, the 2022 rate shock (when this company's book fell from $17.99 to $14.73 and the sector lost 15-25% of book) or the 2013 taper and March 2020 margin-call episodes. We assume book falls 15% in a year, the dividend is cut to $1.40 and then $0.80, book reaches about $8.86, and afterward the company earns only 2.6% on a smaller book. Because the borrowings are the raw material of the business, we do not subtract them separately; the leverage is charged in the bear case. At no point does the bear case exceed book value. What could break this: the repurchase financing is overnight to a few months and must be rolled continuously; in a severe liquidity event lenders can raise margin calls and force sales at the bottom, as happened across the sector in March 2020. The company's large cash balance ($608 million at mid-2026) and Agency-only collateral are real buffers, but a combined rate and spread shock larger than 2022 would leave less than the bear case shows. The January 2026 government directive for Fannie Mae and Freddie Mac to buy $200 billion of Agency mortgage securities supports spreads in the near term, but it is a policy decision that can be reversed and we have not credited it. Margin-of-safety price ($6.50). We are fairly sure of the book value, which is marked to market every quarter, and much less sure of the return on it, which has ranged from minus 10% to plus 22% in single years. The worst plausible outcome is a forced deleveraging that permanently impairs book, and a 7-8 times levered balance sheet funded short-term would survive only by selling assets and probably stock at depressed prices. If we are wrong about the sustainable return and about the depth of the next shock, the shares are worth roughly the bear figure. At $6.50 the dividend, even if cut to the bear case's $0.80 to $1.40, would be 12-20% a year of that price, and a recovery toward a normal return on book implies a value well above that price within a few years; at higher prices a 20% annual return depends on 2025-style spreads recurring. Because book value fell again in early 2026 after a decade of erosion, the margin-of-safety price applies only once book per share has held or risen for two consecutive years with the full dividend paid, showing the dividend is earned rather than funded from capital.