AGNC Investment Corp. (AGNC) Q3 2026 earnings review
AGNC's book value drops 17% and income barely covers dividend
AGNC lost a large slice of its book value in a bond sell-off, while its spread income still covered the dividend. Tangible book value dropped 17% to $7.13 a share after rates rose and mortgage bonds lagged Treasuries. Chief executive Peter Federico said most of the sector's rate damage has already happened and spreads sit near the top of their range. But AGNC ended the quarter with leverage at 7.9 times tangible equity, its highest in two years.
| Tangible book value per share | $7.13 −16.9% from June |
|---|---|
| Net spread and dollar roll income per share | $0.37 against a $0.36 dividend |
| Leverage | 7.9x tangible equity 7.4x in June |
| Monthly dividend | Unchanged at $0.12 |
⚖️ Verdict: 🔴 Bearish
The story got worse because a bond sell-off cut book value, and AGNC came out of it with higher leverage, thinner liquidity and one cent of dividend cover. The good news: hedges now cover 90% of funding, prepayments slowed, and the shares it sold above book value held the loss down. One quarter of slower prepayments does not settle that risk, and the new capital replaced losses rather than adding to the base.
The question now is whether the wider spreads that cost AGNC this quarter become higher earnings or the start of a longer fall. If spreads narrow from here, book value recovers and new purchases earn more; if Treasury volatility stays high, leverage of 7.9 times turns the next move against it into another large loss. Next quarter's net interest spread and book value will tell.
🐂 Bull Case
Stock Sales Above Book Value Softened the Loss
AGNC raises capital by selling new shares a little at a time through an at-the-market programme. Selling above tangible book value helps existing holders, because each new share brings in more cash than the book value it claims. This quarter the mortgage investor sold 97.6 million shares for $992 million, against $167 million three months earlier.
Finance chief Bernice Bell said the shares went at "a significant premium to tangible net book value per common share". The timing helped: S&P added AGNC to its MidCap 400 index during the quarter, and Bell said the rebalancing "drove elevated trading volumes in AGNC's common stock". By our math the average price was about $10.16 a share. The sales kept book value roughly $0.25 a share higher than it would otherwise have been.
But the money replaced losses rather than enlarging the base: by our math the $992 million raised about matched the $983 million fall in stockholders' equity. So AGNC ended the quarter with no more capital than it started with, and 8% more shares.
What to watch: whether AGNC keeps selling shares above the new $7.13 book value, in the next quarterly report. Continued sales at a premium would mean it can rebuild capital without diluting existing holders.
Hedges Now Cover 90% of Borrowings
AGNC hedges its borrowing cost mainly with interest rate swaps, contracts in which it pays a fixed rate and receives a floating one. A rise in rates lifts the swaps as it lowers the bonds. Short Treasury positions do the same job. During the sell-off AGNC added protection rather than cutting it.
- Hedges as a share of funding: 90% at the end of September, from 73% in June
- Net short Treasury position: $11.1 billion, from $2.1 billion
- Gains on swaps and Treasury positions: $3.28 billion
- Losses on mortgage bonds and TBA contracts: $5.31 billion
- Share of the losses the hedges offset: 61%, by our math
The hedges did their job, because rates rose and the swaps gained. The part they did not cover is the part they cannot reach: mortgage bonds fell further than the Treasuries and swaps that hedge them. So AGNC enters the fourth quarter with less exposure to rates and the same exposure to mortgage spreads.
What to watch: the duration gap, AGNC's own measure of rate exposure, in the quarterly filing; it stood at 0.7 years in June. A gap near zero would mean the fourth quarter depends almost entirely on mortgage spreads rather than on rates.
Positives this quarter didn't test
Three parts of the bull case rest on claims the release repeats but cannot yet show in numbers. The market move that would test them had not finished by September 30.
- Supply and demand for mortgage bonds: Federico said "supply-demand technicals remain in balance" with originations low; the quarterly filing's spread table will show whether spreads narrowed once Treasury volatility eased.
- Returns on new money: by our math AGNC added about $2.5 billion of bonds at cost during the quarter; next quarter's asset yield will show whether the wider spreads lifted what it earns.
- Relative value against corporate bonds: no reading in the release; it needs the call or the filing.
🐻 Bear Case
Leverage Rose While Book Value Fell
AGNC owns about $94 billion of government-guaranteed mortgage bonds and funds them with short-term loans called repurchase agreements. Leverage, the ratio of those loans to tangible equity, decides how much a market fall costs shareholders. By our math, a 1% drop in the bonds' value now removes about 8.6% of tangible equity.
Chief executive Peter Federico said rates "increased sharply" in late September and that Agency MBS "was the worst performing sector of the Bloomberg U.S. Aggregate Bond Index during the quarter". Tangible book value fell 16.9% to $7.13 a share. Adding back the dividend, the economic return on tangible equity was −12.7%.
"Despite the challenging macroeconomic environment, there are a number of reasons for optimism with respect to the outlook for Agency MBS once volatility in the Treasury market subsides," Federico said. The balance sheet says something else about risk: leverage ended the quarter at 7.9 times tangible equity, up from 7.4 times in June. Spare cash and unpledged bonds shrank to 57% of tangible equity, the lowest share in five quarters.
This matters because the ratio climbed mostly because equity fell, not because AGNC bought more, so the same portfolio now sits on a thinner cushion. Hedges cover more of the funding than before, but they protect against rate moves, not against mortgage bonds cheapening relative to Treasuries. That relative cheapening is what cost AGNC this quarter.
What to watch: leverage and the liquidity share in the next quarterly report. A ratio still near 7.9 times would mean AGNC is holding its position through the sell-off rather than cutting risk.
Dividend Cover Shrank to One Cent
AGNC pays $0.36 a quarter. It measures what it can afford by net spread and dollar roll income: earnings after funding and hedging costs, before market gains and losses. That figure was $0.37 a share this quarter, down from $0.40 three months earlier, and a cent above the dividend.
The squeeze came from the cost of funding. The average rate AGNC pays on its borrowings, after income from its swaps, rose 0.20 percentage points to 3.09%. The swaps offset less of that cost than three months ago, and asset yields barely moved. So the net interest spread, the gap between what the bonds earn and what funding costs, narrowed to 1.86% from 2.00%.
The other side of the argument is that wider mortgage spreads lift the return on money invested now. Federico said spreads are "approaching the upper end of their multi-year trading range". But that helps new purchases, while the existing book earns what it earns. Another quarter like this one takes the figure below the dividend.
What to watch: net spread and dollar roll income per share next quarter. A figure below $0.36 would mean AGNC pays part of the dividend from capital rather than from earnings.
Prepayment Risk Faded as Rates Rose
Homeowners who refinance pay AGNC's bonds back at par, which hurts when AGNC bought those bonds above par. Prepayments ran at about 13% a year in the first half, the fastest pace in two years. Higher rates changed that, because fewer homeowners can save money by refinancing.
Actual prepayments slowed to an annual 8.2% in the quarter, and AGNC cut its forecast for the life of the portfolio to 6.7%. Federico said "prepayment risk has declined as a result of the increase in rates". The lower forecast produced a one-off $66 million accounting benefit, which AGNC excludes from its spread income.
This eases the concern but does not end it, because one quarter of slower prepayments follows directly from a rate rise that can reverse. The portfolio's cost is now 100.3% of par, so the premium at risk from a return to faster prepayments is small.
What to watch: actual prepayment speed next quarter against the 6.7% forecast. A reading near the forecast would mean the amortisation cost that weighed on the first half of 2026 has gone for now.
Risks this quarter didn't answer
Four standing concerns got no new reading, because the earnings release carries no call commentary and no spread or counterparty tables; those arrive with the quarterly filing.
- Funding concentration: AGNC's own broker-dealer, Bethesda Securities, funded 53% of its repo, or $45.7 billion, and the average loan now matures in 10 days; the filing's counterparty table will show how much sits with the central clearing house.
- Returns on new capital: AGNC has never reported the return it earned on past share sales; any realised figure would move the case either way.
- Mortgage spread versus Treasuries at quarter end: not in the release; the filing's spread table will show whether the market moved outside the range Federico describes.
- Policy direction: the rate on new repo loans was 3.95% at quarter end, up from 3.75% in June; the next Federal Reserve decisions decide whether that keeps rising faster than the swaps offset it.
👓 Other Themes
AGNC Swapped TBA Contracts for Mortgage Pools
TBA contracts are agreements to buy generic mortgage bonds at a future date, and AGNC uses them as a cheaper form of exposure; this quarter they funded 0.13 points below repo. It cut its net TBA position to $2.4 billion from $9.7 billion and bought specific mortgage pools instead. Income from rolling the contracts fell by a third, and the whole portfolio ended at $94.4 billion, 3% smaller than in June.
Rates Rose and Mortgage Bonds Lagged Treasuries
Federico said investor sentiment "turned decidedly bearish" in late September. Rates climbed and volatility jumped, so mortgage bonds lagged Treasuries and became the weakest part of the main US bond index. He argued the damage is largely done, because "most of the duration extension for the mortgage universe has already occurred". Spreads near the top of their range mean higher returns on new purchases if volatility fades.
💲 Other KPIs
Interest income minus interest expense, the GAAP measure, rose 172% from a year ago. The bond book grew and a one-off $66 million accounting benefit lifted the quarter. Interest income alone grew 32%, its fastest rate in five quarters. The figure excludes swap income and dollar roll income, so it understates what AGNC earns on the spread.
Running costs fell to 0.92% of average stockholders' equity on an annualised basis, from 1.12% a year ago. Compensation, the larger line, dropped by a fifth to $16 million. AGNC runs a $94 billion portfolio on about $28 million of quarterly cost, so expenses did not decide the quarter.
🔮 Guidance
Unchanged. AGNC declared $0.12 for October, the same monthly rate it has paid in every quarter we can see. It gives no earnings outlook. The payout runs at $0.36 a quarter against $0.37 of net spread and dollar roll income, so the cushion is one cent. Whether the rate holds depends on that income staying above the payout as funding costs rise.
❓ Key Questions
What return has AGNC earned on the capital it raised?
AGNC sold $992 million of shares this quarter and quotes a return range when it issues, but it has never reported the return earned on past sales. Without that figure the case for issuing cannot be checked.
Did AGNC choose higher leverage, or did losses cause it?
Leverage rose to 7.9 times as equity fell. If management kept the portfolio on purpose to hold wider spreads, that is a view; if it could not sell without realising losses, that is a constraint.
Where was the mortgage spread at quarter end?
Federico says spreads are near the top of their multi-year range, but the release gives no figure. The number sets both the loss already taken and the return on the bonds bought during the quarter.
How has book value moved since September 30?
The sell-off came in the final weeks of the quarter. An update on tangible book value through mid-October would show whether the damage continued or reversed.
At what spread does the dividend stop being covered?
Net spread and dollar roll income is a cent above the dividend. A stated threshold for funding cost or spread would let holders judge the payout from public data.