AGNC Investment Corp. (AGNC)

Business Overview

AGNC is a pure-play agency mortgage REIT: agency MBS, repo funding, interest-rate hedges, and a monthly dividend. Commercial lending, mortgage servicing rights, and origination all sit outside the model, so what the share price is really tracking is one levered bond portfolio.

What an mREIT earned you is the dividend plus the change in book value, and nothing else. On that test FY2025 was a good year: tangible net book value per share rose from $8.41 to $8.88 while $1.44 of dividends were paid, a 22.7% economic return. Market capitalisation was roughly $11.8 billion on 10 June 2026; the last reported tangible book was $8.38 at 31 March 2026.

FY2025 declared dividends totalled $1.44 per share, a ~14.0% yield on the 10 June price, paid monthly rather than quarterly. Income investors treat the schedule as part of the product. It changes nothing about where the money comes from.

How Agency Carry Works

Everything AGNC owns is guaranteed by Fannie Mae or Freddie Mac, so a borrower who stops paying costs the shareholder nothing: the guarantor makes the payment. Strip credit risk out and what is left is interest rates and prepayment. That is the whole reason an agency book can carry seven times its equity in bonds when a commercial mortgage lender runs nearer three. The business is to borrow short in the repo market, hold guaranteed long bonds, and keep the difference.

The difference is thin, and it thinned through 2025. Net interest spread, which counts the swap hedges and the TBAs (forward contracts that carry MBS exposure without owning the bonds yet), averaged 1.92% for the year against 2.42% in 2024. It ran 2.12% in the first quarter and 1.81% in the fourth. A spread that narrow only turns into a double-digit return on equity once it is levered eight times over: Q4's carry, reported as net spread and dollar roll income, was $0.35 a share against $0.36 of dividends declared, on $8.88 of book. The payout is essentially the whole of what the carry earns, which means book value has to grow on bond marks rather than on retained profit.

Leverage is disclosed as tangible net book value at-risk leverage: repurchase agreements, other debt and net TBA positions at cost, divided by total equity less goodwill. That ratio was 7.2× at 31 December 2025, having run 7.5× in March and 7.6× through the middle of the year. Annaly reports 5.6× economic leverage, dividing by total equity including preferred, and the two equity bases land close enough together that the fine print does not account for the gap. AGNC genuinely runs the harder book, which it can because every bond in it is guaranteed. The leverage guide sets the two labels side by side.

The danger in that multiple is not the multiple, it is the repo mechanism underneath it. Lenders hold the MBS as collateral and mark it every day. When prices fall the lender wants cash the same day, and a REIT that has run out of buffer raises it by selling bonds into the market that has just fallen, which turns a paper mark into a loss the book never recovers. Leverage sets how far a price move travels into equity. The margin call decides whether it travels back.

Hedges do not cover this. Swaps and short Treasury positions hedge the interest rate risk, so book value is defended if yields rise across the curve. They do nothing about agency MBS cheapening relative to Treasuries, and that spread is the exposure no agency REIT can insure away. The bad quarters in this sector are usually spread quarters rather than rate quarters, and a hedge ratio near 100% says nothing about them.

Book Value, Prepayment and Economic Return

AGNC reports tangible net book value per share, which takes goodwill and the preferred shareholders' claim out of equity before dividing by the common shares. It rose from $8.41 to $8.88 across FY2025. Add the $1.44 of dividends to that $0.47 of book growth, measure it against the $8.41 the year started with, and you have the 22.7% economic return. The book value guide pairs AGNC's path with Annaly's and explains why the label on the denominator matters when you compare price-to-book across filers.

The same arithmetic runs the other way. In the first quarter of 2026 the monthly cheques still added up to $0.36, but book fell $0.50 to $8.38, and the quarter's economic return was −1.6%. The dividend arrived on schedule and the shareholder ended the quarter poorer. Yield alone will never tell you which of those two quarters you are in.

Prepayment is the other force on book value. AGNC's portfolio prepaid at 9.7% CPR in the fourth quarter of 2025, meaning borrowers were repaying early at an annual rate of 9.7% of the balance, with a projected speed over the bonds' remaining life of 9.6%. Speed matters because agency MBS are negatively convex, meaning the holder gets the worse half of every rate move. When rates fall, borrowers refinance and the bond is repaid at par exactly when an ordinary bond would have risen in price, and the cash comes back to be reinvested at the new lower yield. When rates rise, refinancing stops, the bond extends, and the money is stuck in a below-market coupon. That asymmetry is why agency paper pays a spread at all, and the CPR guide has the mechanics.

Valuation Framework

Price-to-book is the frame, but the question underneath it is whether the return on that book beats what shareholders require for holding a levered bond portfolio. Earn more than the cost of equity and the shares deserve a premium; earn less and book value is the ceiling, which is why so much of this sector trades at or below it. A discount is therefore not automatically cheap, and a premium is not automatically dear. Each is a statement about future returns on book.

The 22.7% is the wrong number to put into that test, because it counts a rise in the value of the bonds that cannot be repeated to order. The recurring engine is spread times leverage. Q4's $0.35 of net spread and dollar roll income annualises to roughly 16% of the $8.88 book, and it was earned at a spread that had already fallen to 1.81%.

Cross-check the yield against the book-value trajectory. $1.44 in annual dividends on $8.88 of tangible book is a high cash return, but if book falls in a rate or spread shock the yield rises mechanically while the total return turns negative. The Mortgage REIT Sector Primer runs its planning case at a 100 bps agency MBS spread over Treasuries and stresses it wider, because wider spreads mark the existing portfolio down long before they lift the income on the next one.

Peer context within agency: Annaly is larger (~$15.9B cap against ~$11.8B) and carries less leverage (5.6× against 7.2×) because it also holds residential credit and mortgage servicing rights, which cannot be funded as hard as guaranteed paper. Hybrid and CRE names in the agency vs hybrid guide take different risks entirely. AGNC is the clean reference point for agency carry.

What to Watch in the Financials

Tangible book per share, beside the dividend. Read the two together or neither means anything. Book holding while the dividend is paid is a business working; book falling while the dividend is paid is capital being returned to you and called income.

At-risk leverage. The 7.2×–7.6× band through FY2025 was the operating envelope. A move towards 8.0× lifts both the earnings and the mark-to-market swing on the same book. Falling leverage is ambiguous: it is prudence if book is steady, and the aftermath of forced selling if book has fallen with it.

The two halves of the spread. The quarterly release splits asset yield from funding cost. The squeeze to watch is funding catching asset yield, which happens when short rates stay high while long rates fall, since repo reprices in weeks and the bond portfolio does not.

Where agency MBS trade against Treasuries. Wider spreads mean better returns on money invested from here and an immediate loss on everything already owned. Which of the two dominates depends on how much dry powder the company has when it happens.

Key Risks

The leverage is real. At 7.2× the portfolio is a little over eight times the equity, so a 1% fall in MBS prices takes roughly 8% off tangible book before any hedge gain. That is the arithmetic behind a −1.6% quarter in a portfolio of government-guaranteed bonds.

Spread risk is uninsured. The hedge book covers rates. If agency MBS cheapen against Treasuries, book falls and nothing in the hedge stops it, and the same price fall is what triggers repo margin calls.

Prepayment. Portfolio speeds near 9.7% are unremarkable, but a refinancing wave resets them fast, and negative convexity means it arrives at the worst moment: premium bonds are repaid at par and the proceeds reinvested at the lower yield that caused the wave.

Paying above book. When the price sits above tangible book there is no cushion between the price and the assets. A quarter of book decline hits the buyer twice, once through book and again if the market decides the premium was not deserved.

Monthly dividend expectations. The market treats the monthly cheque as contractual habit. Any cut hits sentiment harder than at a quarterly payer because income investors structure cash flows around the cadence.