Starwood Property Trust (STWD)

Business Overview

Starwood Property Trust does not own the buildings it makes money from. It lends against them, and it sits above the owner in the queue, so the owner's stake is wiped out before a single dollar of the loan is at risk. That ordering is the business. It writes senior mortgages on offices, flats, hotels and warehouses, adds junior loans that rank behind the senior debt but ahead of the owner, and runs three further arms alongside the lending: infrastructure loans, a net lease property portfolio, and a servicing business that works out troubled commercial mortgages for other people's securitisations. Diversity is the pitch. Blackstone Mortgage, the obvious comparison, does only the first of those things.

At 10 June 2026, 370.6 million filed shares implied a market capitalisation near $6.38 billion against GAAP book value per share of $18.34 at 31 December 2025.

FY2025 distributable earnings were $615.5 million, or $1.69 per diluted share, against a declared dividend of $1.92 ($0.48 a quarter, unchanged for over a decade). That is 0.88× coverage: the company paid out 23 cents a share more than it earned on its own preferred measure. The dividend has been held at $0.48 through the first half of 2026, on quarterly distributable earnings still running below the rate needed to fund it.

How CRE Lending Economics Work

The income is interest and fees on loans. Commercial lending was $16.6 billion at the year end, infrastructure lending $2.8 billion, and residential loans held for sale another $2.3 billion, which is $21.2 billion of loans once reserves are taken off. Undepreciated assets, meaning the whole company including the buildings it owns outright and the servicing platform, came to $30.7 billion.

What stands between the lender and a loss is the borrower's own money. Write a loan at 65% of a building's appraised value and the value has to fall by more than a third before the loan itself is short. But that cushion is measured against an appraisal struck on the day the loan was written, and appraisals do not update themselves. An office building valued in 2021 and half empty in 2026 supports a loan-to-value that stopped being true years ago, which is why the origination figure and the current figure are different measurements wearing the same label.

Carrying credit risk is also what caps the leverage. Starwood runs debt at about 2.4 times undepreciated equity. An agency mortgage REIT runs at seven or eight times, because government-guaranteed bonds can be funded at a small haircut and sold in an afternoon; nobody funds a book of office loans on those terms. The gap is the price of the credit risk, not a difference in nerve, and the leverage guide works through why the two numbers cannot be ranked against each other.

Office exposure is where the comparisons go wrong. Starwood reports US office at 8% of its undepreciated asset base, its lowest ever share, and that denominator is the entire $30.7 billion company. Blackstone Mortgage reports office at 27% of net loan exposure, dividing by a far smaller number. The same dollars of office produce very different percentages, so 8% against 27% is not a ranking. Convert to dollars before comparing, as the CRE debt guide sets out.

Roughly $1 billion of commercial loans sat on non-accrual at the year end, meaning Starwood had stopped booking interest on them, and a further $624 million had already been foreclosed and taken onto the balance sheet as property. Against that stress the company carried $680 million of reserves, $480 million of expected credit losses on loans still held and $200 million of write-downs on the foreclosed property. That is $1.84 a share against an undepreciated book value of $19.25. Two cautions on those numbers. A credit loss reserve is management's estimate of a loss it expects, not one it has taken, and it moves in both directions. And a loan whose maturity has been extended or whose terms have been softened stays performing on paper, so the accruing balance is only reassuring once you know how much of it got there by being given more time.

Book Value and Dividend Coverage

Two book values are disclosed and they measure different things. GAAP book value per share was $18.34; undepreciated book value per share was $19.25. The 91 cent gap is accumulated depreciation on the buildings Starwood owns rather than lends against, a bookkeeping charge that does not describe what those buildings would fetch. Use the GAAP figure for a price-to-book screen, because that is the basis the rest of the sector is quoted on, and keep the undepreciated one for judging how much reserve is sitting inside the equity. The book value guide sets out the wider frame.

Distributable earnings is the company's own cash-basis measure, and it is the line the dividend is declared against: GAAP net income with depreciation, unrealised marks and non-cash pay stripped out. FY2025 distributable earnings of $1.69 a share against the $1.92 dividend leaves a shortfall of about $85 million on the current share count. A shortfall of that size can be funded from liquidity or new borrowing for a while. Run for long enough it is being paid out of book value, and the choice is between cutting the dividend and letting the equity erode.

Where the collateral itself reprices matters more than any single quarter's coverage. The cap rates guide gives the property-type evidence for stress-testing those marks, and non-accruals of about $1 billion against an 8% office share are a reminder that the problem loans are not all offices.

Valuation Framework

A discount to book at a lender is neither a bargain nor a warning by itself. It is the market's estimate of how much further book value has to fall, so the question is how it compares with what management has already put aside. The discount to book was worth less than the reserves already inside book ($1.84 a share). On that reading the market was pricing rather less further damage than Starwood has itself provided for, which is a different position from Blackstone Mortgage's.

Two things sit permanently against the multiple. Starwood is externally managed by Starwood Capital, and the manager takes 1.5% of equity a year plus an incentive fee, $137 million in 2025, roughly 37 cents a share against $1.69 of distributable earnings. That charge does not go away in a bad year, and the market prices it into the discount. The board authorised a $400 million repurchase programme in February 2026 and had used $30 million of it by the end of June: buying stock below book adds to book value per share, which is the one lever available while coverage is short.

Set against Blackstone Mortgage, the office percentages tell you nothing and the reserves tell you a good deal. Starwood carries $1.84 a share of reserves inside a $19.25 undepreciated book, about 10%; Blackstone Mortgage carries $1.76 inside a $20.75 book, about 8%. Two points apart, on credit disclosure that looks nothing alike. What separates them is what reached the earnings line: Starwood's distributable earnings stayed positive and short of the dividend, while Blackstone Mortgage's turned negative after charge-offs and the dividend was paid anyway. The market prices the two credit lenders close to each other.

The agency names are not comparators. Annaly and AGNC hold guaranteed paper that is marked daily and carries no credit risk at all, which is why they trade at or above book while credit lenders trade below it. Rithm is a hybrid holding mortgage servicing rights and an origination business, covering its dividend 2.35 times on its own measure. Stay inside the CRE lenders for relative value.

What to Watch in the Financials

Distributable earnings against the dividend. Coverage of 0.88× is the anchor. Earnings climbing back toward $1.92 a share is the only route to a covered payout that does not involve cutting it.

Non-accrual trajectory. Watch whether the billion-dollar balance is resolving or growing, and at what price the resolutions happen. A loan returning to accrual adds back interest income; a new one hits book value and distributable earnings in the same quarter. Management has said it expects to resolve close to $900 million of underperforming assets by the end of 2026, which is the claim to hold it to.

Office collateral performance. US office is 8% of the undepreciated asset base. The number worth tracking is whether office accounts for a disproportionate share of the non-accrual balance, because that says whether the small percentage is doing the damage.

Book value per share. Each quarter's move absorbs new provisions and realised losses, and buybacks below book push the other way. The two things that move the price-to-book screen are whether book stabilises and whether coverage crosses 1.0×.

Key Risks

Dividend coverage below 1.0×. Distributable earnings of $1.69 a share did not cover the $1.92 dividend in 2025, and the first half of 2026 has not closed the gap. A second full year short forces the choice between the payout and the book value.

Reserves are an estimate, not a settled bill. The $680 million already provided is management's own view of what these assets will lose. Softer collateral values raise it, and the increase lands on book value and distributable earnings together. Resolutions above carrying value release it back the other way.

Extensions flatter the performing book. Modifying a loan or pushing out its maturity keeps it accruing, so the share of the portfolio described as performing depends partly on how much of it has been given more time rather than repaid.

Funding and maturity walls. A $21.2 billion loan book has to be refinanced continuously, and the leverage is modest precisely because the collateral cannot be sold quickly. If secured funding tightens, rolling the book gets harder before credit losses even accelerate.