The Latticework Letter — Issue #015

Something strange happened this week, and if you only watched the Fed odds you missed it. August PCE landed soft on Tuesday — headline up 0.3% against a 0.4% forecast, core up 0.2% against 0.3%, core 2.9% year-over-year, in line — and futures did what futures do: the probability of an October hike fell from roughly 51% to about 38%. Last week the market put three-in-four odds on October. This week it’s closer to one-in-three, with the pressure rolling to December. The Fed, in other words, got friendlier. And the 10-Year Treasury responded by… rising all week — from Friday’s 5.17% close through 5.21% Monday, 5.26% Tuesday, and roughly 5.33% Thursday morning, levels last seen in the summer of 2007, up more than half a point in a month.

Read that again, because it is the most important macro fact in net lease right now: the long end of the curve went up while Fed-hike odds went down. When yields rise because hikes are coming, a pause fixes it. When yields rise while hike odds are falling — on supply, on term premium, on the suspicion that inflation is structural rather than cyclical — there is no cavalry. Friday’s September jobs report made the point for us: 29,000 jobs against roughly 90,000 expected, the prior months revised down, unemployment up to 4.2% — and the market took the October hike off the table entirely, rolling its one remaining 2026 hike to December. The 10-Year’s reward for the weakest print of the year? A few basis points, back to about 5.28%, still within sight of its 19-year high. A labor market that soft buying that little relief is the term-premium story in a single trade.

Oil, at least, blinked. Monday opened ugly — Trump rejected Iran’s conditional offer to reopen the Strait of Hormuz (their price: lift the naval blockade and oil sanctions), WTI jumped about 4% to above $95, Brent touched $108 — but the physical market kept healing anyway. Saudi Arabia resumed tanker loadings at Yanbu, Hormuz flows recovered to 13.2 million barrels a day, near pre-war levels, and by Thursday WTI sat near $91 and Brent around $101. Your gas-station tenant finally caught a break: AAA’s national average slipped to $4.41, down about seven cents from $4.48 a week ago — though still 32 cents above a month ago, and diesel ($6.39) is barely off the all-time record $6.53 set September 22. Remember the escalator rule from Issue #014: retail prices follow crude up the staircase and down the escalator. Seven cents in a week is the escalator working.

Put the week together and the arithmetic lands in one place. Boulder’s Q2 average asking cap rate — still the freshest broad private-market print — is 6.82%. Against the 5.28% where the 10-Year ended the week, the average net-lease deal now offers about 154 basis points over the risk-free rate. It was 170 last week. It was roughly 240 when those Q2 asks were set. The public market keeps marking the repricing in real time — Realty Income is down about 10.7% in a month (Scotiabank cut it to Sector Perform on September 23, price target $67 to $59), Essential Properties is off 11.4%, NNN REIT 8%, Agree 7.2% — while the private market’s quote sheet still hasn’t moved. And August’s headline transaction tape — $107 billion, up 127% — was two-thirds entity-level M&A; strip it out and single-asset sales, the market net lease actually trades in, fell 21%. Volume first, then price. Every week the Treasury holds above 5, the appraisal gap gets harder to ignore and easier to measure.

The largest net leases being signed in America right now are not drugstores or dollar stores. They are data centers, and the week’s quietest big number is the best way in: on September 25, CleanSpark — a bitcoin miner pivoting to AI infrastructure — closed $2.276 billion of senior secured notes at 7.875%, issued at 98.5, maturing October 1, 2031, secured by first-priority liens and a parent completion guarantee, to finish building its Sandersville, Georgia campus. A miner borrowed two and a quarter billion dollars, in this rate environment, against a building and a lease. To understand why lenders said yes — and what it has to do with your Walgreens — look at the machine Hut 8 built first.

Start with the structure, because it will look eerily familiar. Hut 8’s River Bend campus in Louisiana carries a 15-year lease with Fluidstack for 245 MW of IT capacity: $7.0 billion of base-term contract value, triple-net, with 3.0% annual escalators and three five-year renewal options that take the potential total to $17.7 billion — expected net operating income averaging about $454 million a year. The tenant is a young GPU-cloud company you have never underwritten, and you don’t have to, because Google (AA+/Aa2) backstops the lease payments and pass-through obligations for the entire 15-year base term. The deal sits inside a multi-gigawatt partnership with Anthropic announced last December; first data hall delivers in Q2 2027. Now say it in net-lease English: a 15-year absolute-NNN, credit-tenant lease with 3% bumps and a guarantor better-rated than any tenant on Boulder’s survey. The AI industry did not invent a new instrument. It picked up the oldest one in your drawer — the long-dated net lease with a credit wrapper — and scaled it by three zeros.

The capital markets are already pricing the familiar shape accordingly. Hut 8’s Beacon Point project in Texas — 352 MW leased for 15 years at $9.8 billion of base value to a confidential tenant rated AA− or better, same 3% escalators, options to $25.1 billion, roughly $655 million of average annual NOI — financed itself in June with $4.25 billion of non-recourse senior secured notes at a 6.129% coupon, priced at Treasuries plus 165, rated Baa2. Treasuries plus 165 is investment-grade net-lease pricing. Across the two campuses, a company that mined bitcoin three years ago has raised $7.5 billion of investment-grade construction financing against roughly $16.8 billion of contracted triple-net, take-or-pay leases. Meanwhile CleanSpark — no hyperscaler backstop, miner balance sheet, completion risk — pays 7.875%. There is your credit spectrum, same as it ever was: the Google-backstopped lease prices like a Chick-fil-A ground lease; the sponsor-credit deal prices like casual dining. The wrapper is identical. The spread is the underwriting.

So what does a 2,000-square-foot-store investor do with this? Three things. First, understand where the marginal institutional dollar is going, because it is not going to your auction: a lender or fund that can buy Baa2 paper at T+165 against a 15-year Google-backstopped NNN lease needs a reason to take retail re-leasing risk at 154 over, and that reason is called a wider cap rate. The competition for “long-duration credit net lease” capital now includes gigawatts. Second, respect what is not transferable. Your Wawa sits on a corner that has value if Wawa leaves; it can become a taqueria, a clinic, a car wash. A 245-MW data hall is the purest single-tenant, single-purpose real estate ever built — the residual value without the tenant is a question nobody has had to answer yet, which is precisely why every one of these leases is take-or-pay with an investment-grade backstop. They need the credit because the building is the credit. Third, notice the vintage risk running the other direction: those 3% fixed escalators are being signed into a world of 3%-handle core PCE. The AI landlords are making the same bet 2021’s net-lease buyers made — that fixed bumps beat future inflation. If you sold flat-lease drugstores in 2022, you know how that story goes.

Leslie’s — the 63-year-old, ~1,000-store pool-supplies chain — filed a prearranged Chapter 11 in the Southern District of Texas on Wednesday. The headline terms: 76 stores closed with the filing, more than 900 stay open, the plan eliminates about $685 million of funded debt — roughly 90% of it — and hands majority ownership to lenders holding over 80% of the funded debt. Existing equity is canceled (Nasdaq suspends LESL trading October 6), the case is funded by a $90 million new-money DIP, a $225 million ABL facility, and a $60 million backstopped equity raise, and emergence is targeted for early 2027. For landlords, the operative sentence is in the company’s own release: Leslie’s “will continue to evaluate its real estate portfolio,” with Hilco Global engaged to manage it. Translation: 900-plus leases now go through the Chapter 11 sorting machine — assumed (cure paid, lease continues), assigned, or rejected (claim capped at the greater of one year’s rent or 15%, up to three years). A prearranged case with lender support and a 100-day confirmation target is the orderly kind — but orderly is a description of the process, not of your site’s odds. If you own one, the triage is the same as ever: store-level sales if you can get them, your location’s rank in the market, and your downside re-let math, run now rather than at the rejection notice.

America’s Car-Mart bought another week — its sixth. An 8-K filed Wednesday extended the lender waivers to October 8, covering the existing and anticipated defaults, the minimum-liquidity relief, and the collateral-coverage test, while a special committee of the board evaluates “strategic alternatives” — financing or restructuring — in talks the company describes as active. The chronology now reads September 7, 11, 18, 24, October 1, October 8. The read is unchanged from Issues #013–#014: a lender extending a week at a time is a lender close to a signature, and the new wrinkle — a formal special committee shopping alternatives — says the signature is being negotiated now, not later. The going-concern math hasn’t moved (Q1 revenue down 57.3%, a $69 million quarterly loss), the plaintiffs’ bar (Rosen, Cruz, Smith) is still circling, and for landlords of the ~90 operating dealerships the posture is still store-by-store triage: the dirt, not the drama. Next cliff: Wednesday, October 8.

Salad and Go got its answer Tuesday: the bankruptcy court approved 7 Brew’s bid — now $123 million for roughly 60 leases, against the $143.2 million, 73-site deal struck at auction. Read the delta, because it’s a net-lease lesson in one line: something like ten leases fell out between auction and approval, reportedly where landlords did not consent to assignment or sites didn’t fit (36 Arizona, 19 Texas, 3 Nevada, 5 Oklahoma made the cut). Assignment provisions you negotiated years ago just decided whether your box gets a growing coffee tenant or goes dark into the estate. Two grace notes: the sale is expected to pay unsecured creditors in full — genuinely rare — and 7 Brew (800-plus stands, aiming for 1,000) converts the boxes to drive-thru coffee “over time,” which is the Issue #013 supply thesis arriving on schedule.

Briefly noted: Starbucks shuttered about 250 North American stores this week in its second closure wave, taking ~$300 million of charges — company-operated urban boxes mostly, but every closure list that touches freestanding drive-thrus deserves a read if you hold one. Realty Income’s €528 million KKR joint venture closed September 30 — more on that below. And the OZ 2.0 nomination deadline passed Monday with 16 of 56 jurisdictions filed, about 1,735 tracts (~26% of the expected 6,575); Texas submitted its full 608. The other 40 jurisdictions — including California (620-tract cap), New York (429), and Florida (342) — run to the October 28 extension, and all designations still take effect January 1, 2027.

On the calendar: Boulder’s Q3 market-wide report — the first broad private-market print taken entirely above a 5% 10-Year — is due any day. Car-Mart’s sixth waiver expires Wednesday, October 8. Getty and Essential Properties both report Q3 on October 21. OZ 2.0 extension deadline October 28. FOMC: October 28 — October is now priced out; December is the live meeting.

That is the share of Realty Income’s first-half 2026 investments funded by selling its own stock. The historical average is about 47%.

Hold that number up against the transaction that closed September 30: Realty Income’s first euro-denominated joint venture, in which KKR-advised capital paid €528 million for 49% of a 54-property portfolio across Spain, Ireland, Poland, and the Netherlands — roughly €67.7 million of projected first-year cash NOI, about a 5.9% initial cap after management fees, a 7.2-year weighted average lease term, 59% investment-grade rent, with Realty Income keeping 51% and the keys. KKR’s targeted return is 6.3–6.5%. Why would the sector’s blue chip sell half of a stabilized portfolio at a 5.9? Because the alternative funding source is its own equity, and with the stock down double digits and yielding 5.7%, issuing shares to buy 7-caps no longer makes the spread math work. Public equity funded 18 cents of the first-half dollar; something has to fund the other 82, and this week it was Henry Kravis’s heirs.

This is the same film we watched in the Spotlight, projected onto a different screen. Hut 8 funds growth with non-recourse bonds against Google-backstopped leases; Realty Income funds growth with private-capital JVs against euro-denominated ones; J.P. Morgan raised $1.1 billion to buy the sale-leasebacks CFOs won’t sell yet; CBRE IM bought the platform that originates them. Everywhere you look, the equity market’s door is half-closed and the structured-capital door is wide open. For the private investor, two implications. One: the REITs are not leaving the bid — they are changing whose money they bid with, and private-capital return targets (KKR’s 6.3–6.5% levered, in Europe) tell you where institutional pricing actually sits. Two: when the cheapest capital in the sector starts renting other people’s balance sheets, the expensive capital — yours — should demand more spread, not less. 149 basis points, for reference, is less.

Munger kept a short list of ways smart people ruin themselves, and near the top was the circle of competence — not the size of it, which he said matters far less than people think, but knowing where the edge is. “Knowing what you don’t know is more useful than being brilliant.” The corollary he loved: the most dangerous ideas are not the foreign ones, which trip your alarms, but the familiar-looking ones, which don’t.

Now reread the Spotlight with that lens. A 15-year lease, triple-net, 3% bumps, investment-grade guarantor — every pattern-matching instinct you’ve built over twenty years of Walgreens and Wawa says I know this animal. That recognition is exactly the hazard. The form is inside your circle; the substance — power procurement, GPU depreciation curves, whether a 245-megawatt shell has any second user, whether the guarantor still wants the capacity in year eleven of fifteen — is not, and no amount of fluency in cap rates converts into fluency in compute. The 2007 vintage of this mistake was CMBS buyers who knew bonds and therefore assumed they knew mortgage credit; the 2021 vintage was net-lease buyers who knew drugstores and therefore assumed a 4-cap flat lease was conservative. The instrument was familiar. The risk was new. Munger’s test for the edge of the circle was brutally practical: can you name the two or three factors that will actually determine this investment’s outcome, and would an expert in that field agree with your list? For a dollar store, you can — credit, corner, re-let. For a data hall, be honest about whose list you’d be borrowing. None of which means the AI net lease is a bad deal — Google’s guarantee is realer than most tenants’ balance sheets. It means the 7.875% and the 6.129% in that story are not two prices for the same risk; they are the market telling you how much not-knowing costs per year. Inside your circle, you get paid for judgment. At the edge of it, you pay — and the first installment is usually invisible, because everything looks so wonderfully familiar.

The Latticework Letter is published weekly on Friday mornings. It is independent analysis — we have no brokerage relationships, no listings to push, and no financial products to sell. Our only interest is giving you a clearer picture of the market.

Forward this to someone who owns NNN assets and is tired of getting their intelligence from people with something to sell.