The refinancing wall is breaking in layers

A Boston tower split $856 million into two financings last week, and the way the money divided tells you more about the direction of commercial real estate credit than any single loan. Winthrop Center put $281 million of C-PACE on its condominium portion and $575 million of securitized debt on its office portion, one leg running 20 years and the other five. The refinancing wall the market spent two years bracing for is turning out to be a sequence of layers, each pricing a different risk.

PWD's tracking shows the same pattern across the week's debt and deal coverage: a small-balance bank loan on vintage multifamily, a construction mortgage splitting apartments and office, a convertible bond on data centers under construction, and two office financings in the data — the Boston tower and a $147.5 million Mountain View package pairing C-PACE with senior debt. JLL's July indices put lenders at a record reading and buyers at a one-year high, the spread between the two series—the bid-credit gap—closing even as July's CRE volume hit a two-decade high only because $33.8 billion of data-center deals propped up an otherwise flat market. Strip out data centers and the recovery is exactly what the layers suggest: narrow, specific, and priced one asset at a time.

At the safest end of the stack, patient capital is taking stabilized trophy assets: in Boston, the C-PACE piece sits against condominiums inside a mixed-use tower whose exit is a sale to an owner-occupier rather than a lease roll, and in Mountain View a fully leased, newly built office building refinanced ahead of maturity with a package pairing C-PACE and senior debt. C-PACE extends a stabilized asset's maturity without forcing a sale into a thin market. Trophy-office capital is back, and it arrives in layers.

The bid side is moving to meet that capital: in Phoenix, a Transwestern/C-III joint venture paid $86 million for Esplanade III, pricing 17% vacancy and a 4.4-year lease roll into a basis up 43% from the previous trade, while Petraville's $212 million purchase of an Euljiro office tower in Seoul put a $538-a-square-foot price on the rent roll with the seller unnamed and the income doing the talking. Those are price discoveries, and they let the next layer of capital move.

One rung down, securitized debt is pricing office cash flow: the $575 million Winthrop office piece is a claim on rent with the exit value left to the equity holder, and office securitization, the slowest channel to reopen, has returned as the clearest sign that the bid-credit gap is narrowing. The same flight to cash flow shows up in Canada, where CoStar's read makes high-end offices the defensive trade and luxury apartments the vacancy risk. The office trade has narrowed to a specific bet on buildings that produce rent today.

Construction capital is the middle of the stack, and the week's deals show lenders splitting risk within a single asset rather than avoiding it: Bank OZK's $165.9 million mortgage on Southern Land's 2811 Kirby in Houston pairs a 38-story apartment tower with a 10-story office block, an underwrite that has to believe the two uses will not peak and trough together. The same lender is behind a $75.5 million construction loan on two Norwalk, Connecticut office buildings that a joint venture plans to convert into 286 apartments, making the building's next use the collateral and leaving the current rent roll out of the calculation. A lender willing to underwrite that conversion is the bid-credit gap closing in its most literal form.

At the small-balance end of the construction and ownership spectrum, local banks are writing patient checks on vintage multifamily. Bascom refinanced the 133-unit Charleston Apartments with an $8.5 million bank loan, a dollar amount small enough to miss but a tell too loud to ignore: neighborhood risk is being priced and held by balance sheets that understand it, without a syndication desk or a bond market. Rockpoint and Holland are making the same patient bet on 2029 San Jose rents with a 311-unit Stevens Creek development, and LeClaire Courts in Chicago closed a $127 million financing as Ogden Commons reached full lease-up.

At the riskiest end, the market is asking bondholders to take construction risk on assets that do not yet produce income. Nebius, the Nvidia-backed neocloud, raised $5.75 billion in convertible debt to fund construction and GPUs, lifting its total convertible debt to $12 billion against $529.8 million of revenue. There is no stabilized cash flow to underwrite; the collateral is the promise that AI demand will fill the racks. The data-center theme ran through the whole week: July's record volume was a data-center story, and Mapletree's 22-building, 3.1 million-square-foot sale will test whether buyers pay for stabilized data-center income or still price the asset class as a construction story.

Stabilized assets clear first, then cash-flow assets, then construction, then assets whose revenue is still being built. A C-PACE bondholder and a bank lending on vintage apartments draw on different capital markets even when they appear in the same week's news, each making a separate decision about a separate maturity. No single credit event has to carry the market because the market is a stack of smaller agreements.

Industrial capital is clearing at both ends too: EQT exited 10.5 million square feet of Southeast logistics to LBA in a $1.2 billion rent-reset purchase, the industrial bid finding a price at scale, while the development pipeline is being built by balance-sheet capital—a $20 billion land bank outbuilding private equity across the Sun Belt and Hines mapping a $1 trillion advanced-manufacturing wave to six metros. Those are supply-side bets on where the next layers of demand will form.

The market is dismantling the refinancing wall in tranches rather than waiting for it to break, and the allocation that follows is to match capital to the specific layer: C-PACE on a stabilized condo carries a different risk from a convertible bond on a data center under construction, and the market is finally pricing them accordingly. The market has stopped asking when the wall breaks and started asking which layer it can price. The next test is Mapletree's sale, and whether buyers pay for stabilized data-center income or still price the asset class as a construction story.