Realty Income

On 30 June we wrote that Realty Income is a landlord of the finest, dullest kind, that it must be read on AFFO and not on the fifty-one-times "P/E" the screens print, that its virtue is a dependable monthly cheque rather than capital gains, and that the thing to do was own it for income at $62 and add below the mid-fifties, where the yield tops six percent. Ten weeks, one earnings report, one rating and one dividend increase later, here is the ledger.

One of seven wrong, and it is the interesting one. We described a REIT that "cannot fund itself" and grows only by issuing shares. That was true for thirty years and it stopped being true this year: in 2026 the company has funded roughly four-fifths of $10 billion of investment without selling public stock, by selling stakes in its own buildings to Apollo, by drawing a sovereign fund's money into a logistics venture, and by running an open-ended fund that now manages $3.3 billion of which it owns less than a quarter. A widely followed REIT analyst who met the chief executive in Milan this week is calling it "the massive transformation nobody saw coming." We would put it more plainly: the landlord became a fund manager, the direction was visible in June, and the surprise is the speed. That is why Part V of this report is a drawing rather than a table. Everything else — the rent, the cover, the rating, the sensitivity to the bond market — happened exactly as described, which is why the score does not move.

Before the transformation, the thing being transformed. Realty Income owns 15,588 buildings in all fifty states, the United Kingdom and eight countries on the Continent — the most geographically spread portfolio on our board and, since June, the one whose non-American share has grown fastest. The picture below is the 10-K's state-and-country table drawn as a tile grid, the same form we used for NNN so the two can be compared: NNN's rent is 18% Texas; Realty Income's is 9.6% Texas and 14.3% United Kingdom, which out-rents every American state.

Three things the drawing says that the tables do not. First, the concentration is low and getting lower: the largest state is under ten percent of rent, the largest client (Dollar General) is 3.3%, and the largest industry (grocery, 11.1%) is the one least likely to be disrupted by a smartphone. Second, the map is now two maps: a fifth of the rent comes from outside the United States — the UK at 15.0% and continental Europe at 5.5% at 30 June — and roughly a third of 2026's new money went there, funded with euro bonds at 3.6%, because a British supermarket at a 7% cash yield financed at 3.6% is a spread the U.S. no longer offers. Third, what it owns is changing at the edges, not the core: retail is still 78% of rent and 14,913 of the buildings; industrial has grown to 16%; and the two casinos and the data-centre venture live in the 5.5% labelled gaming and other. The occupancy line — 98.8%, never below 96% at a year-end — is the reason the dividend has been paid 675 months in a row, and it did not move this quarter.

Nothing about the method changed since June; only the inputs did. Read on AFFO, the dividend consumes 74.5% of what the buildings actually earn, and the shares trade at a little over thirteen times that figure — a full turn cheaper than ten weeks ago for a company whose guidance went up. Our feed also prints an Altman-Z of 0.9 (the "distress zone") and negative owner earnings of −$0.85 a share: both are the model treating a landlord like a factory, exactly the faults we named at NNN and Rexford. Discard them. The numbers that matter for a net-lease REIT are the four in Part VI: the initial cash yield on what it buys, the cost of the money it buys with, the occupancy, and the credit of the tenants.

Every REIT on our board gets a picture of what it owns. Realty Income's June report described the portfolio; this one draws the thing that changed. The upper panel is the funding model — the share of investment paid for with newly issued public stock — and the fund the company is steadily selling its own properties into. The lower panel is what 2026's money actually bought, block width proportional to dollars, and the number that matters written on each: the initial cash yield.

The upper panel is the good news. For three years public stock paid for about half of everything Realty Income bought; this year it has paid for less than a fifth. The difference came from three places, all new since the spring. In March, Apollo's insurance funds paid $1.0 billion for 49% of a venture holding 492 existing retail properties — Realty Income keeps 51%, keeps the management, and can buy Apollo out between years seven and fifteen at a return capped at 6.875% (approx. — company release). In January, GIC committed to a build-to-suit logistics venture of more than $1.5 billion, which also carried the company into Mexico. And the US Core Plus Fund — an open-ended vehicle that Realty Income seeded with its own buildings in September 2025 — has taken in enough institutional money that the company's stake fell from 38.5% on 1 January to 23.6% on 1 July, returning roughly $600 million of its own equity while it kept the management fee (approx.). Fee-earning assets went from $0.6 billion to $2.7 billion in two quarters. The fee income is still a rounding error — about $3 million a quarter against $1.1 billion of quarterly AFFO — but the equity it replaces is not.

The lower panel is the part the transformation cannot hide. The headline cash yield on 2026 investments is 7.3%, which looks as good as Agree Realty's 7.0% or NNN's 7.3%. It is a blend. Pure real-estate acquisitions — the thing Realty Income has done for fifty years — came in at 6.4% in Q2, down from 6.7% in Q1, because private buyers with cheap money have compressed cap rates on exactly the buildings it wants. What lifted the blend to 7.3% was $1.7 billion of loans and structured credit at 8–9%, from a company whose chief executive said on the same call, "We are not a lender. We are not a bank." He is right that it is not a bank. It is a landlord that, to keep its headline spread, is lending money at nine percent to other people's real estate. The two panels are the same story from two sides: the equity got cheaper and the buildings got dearer, and the company is using the first to keep buying despite the second.

The dividend is as safe as any on our board and the honest finding is, again, in question six. The raise is five-hundredths of a cent, three times a year — 1.9%, while AFFO grows 4%. Read charitably, the board is building a cushion for the fund-and-data-centre era; read plainly, it has quietly told you that the cheque will grow at about the pace of a lease escalator. Set that against NNN's +3.3% raise this summer and Agree's monthly growth in the same range, and the "Monthly Dividend Company" is the slowest-growing dividend in the peer group we cover. It is the price of size. It was the price of size in June, too.

In June the central question was rates, and it still is: the price fell four percent because the ten-year rose fifty basis points, and nothing the company did in the quarter mattered to the tape. But the quarter raised a second question that will outlast this rate cycle, and it is the one the Milan analyst is excited about: if a REIT's growth no longer depends on selling its own stock, what is it?

The bull case, stated fairly. A net-lease REIT's growth arithmetic is volume × spread ÷ dilution. For thirty years Realty Income's spread was the gap between a 7% building and a ~6% blended cost of stock and bonds, and its dilution was the stock it had to sell to buy the building. This year it has found equity that costs less than its own: Apollo's money at a 6.875% cap, a sovereign fund's, and a private fund's, on which it also earns a fee. The company's stated ambition is an "ecosystem" that "maximises the utilisation of a platform with the lowest cost of equity capital" — and a $15 trillion addressable market on its slides that now includes a trillion of data centres, of which it bought 45% of a $6 billion, 400-megawatt Northern Virginia venture the day our June report went out (approx. — company). If the fees scale and the equity stays cheap, the growth rate goes up and the multiple, in time, follows: asset managers trade at twice a REIT's multiple.

The bear case, stated fairly. The fees are $12–13 million a year against $4.2 billion of AFFO — three-tenths of one percent — and no analyst has changed a target for them. Apollo's 6.875% is not obviously cheaper than 3.9% debt or than the equity Realty Income already retains. The fund lets the company buy at 5.8% cash yields that would be dilutive on its own balance sheet, which is another way of saying it is buying things it should not buy with its own money and charging a fee for the privilege. Managing outside capital alongside your own creates allocation conflicts the company has not been asked about in public, and its general counsel's chair has been empty since 2 September. And the transformation does not touch the number in Part V: 6.4% on a building is 6.4% whoever holds the equity.

Our read: the transformation is real, early, and priced at zero. The market has not paid a cent for the fund, the fees or the data centres — the stock is a full turn cheaper on AFFO than in June — and it has not charged for the conflicts either. That is roughly right for now: the fees are too small to value and the risks too new to price. What the pivot has already done is improve the one variable we said in June could stall the machine, the cost-of-capital trap: a REIT that can fund four-fifths of its growth without selling stock at $59 is a REIT whose growth no longer stops when its share price falls. That is worth something, and it is worth exactly the growth dial staying at 5 instead of slipping to 4 on a 1.9% dividend raise and a 6.4% cap rate.

The read-across is the same as in August and it has sharpened: the pure-plays a tenth of Realty Income's size are compounding AFFO per share at 6–8% on 7.0–7.4% cap rates without leaving retail, and Realty Income is compounding at 4% on 6.4% cap rates with a third of its money in loans. Size is the reason — a $4.2 billion AFFO base needs $10 billion of volume a year to move four percent, and there are not $10 billion of 7% dollar stores for sale — and size is also the reason the private-capital model exists: it is the only way a company this large can keep the spread. Whether that makes it the best net-lease REIT to own is a different question from whether it is the safest, and the answer to the second is still yes.

Capital allocation is the story of the quarter and we score it up, not down: the company found three sources of equity cheaper than its own stock, refinanced at 3.9%, and raised the dividend — while committing $1.4 billion to a sector it has not yet proven it understands, and doing so with the legal chair empty. Dial: 7, unchanged. The pivot earns a point; the data-centre bet and the vacancy give it back.

The valuation of a bond-like REIT is a spread, and the spread tells two stories at once. Against the dividend, Realty Income now pays only about sixty basis points more than a ten-year Treasury, the thinnest cushion in years and the reason a column of "does the monthly dividend still make sense?" pieces appeared this month. Against AFFO — the earnings the dividend comes from — the spread is 260 basis points and has not moved since June, because the price fell exactly as fast as the bond rose. The first spread is what a retiree comparing cheques sees; the second is what an owner sees, and it says the company got a turn cheaper on a number that went up. Our own arithmetic is unchanged in method and lower in level: 15× a $4.45 AFFO is $67, which is also the street's mean; 14× is $62; and the point at which the yield crosses six percent — where Mr. Market pays you for the rate risk rather than merely charging you for it — is ~$54. In June we said "add below the mid-fifties." The price has done two-thirds of the journey.