Eight hundred and seventy-five billion dollars of commercial mortgages mature this year, and most of them will be met with an extension rather than a payoff. The extension is where a non-recourse loan quietly stops being non-recourse.
The Mortgage Bankers Association puts 2026 maturities at $875 billion, about 17% of the $5.0 trillion outstanding—down 9% from the $957 billion that came due in 2025, but still one of the heaviest years on record. Thirty percent of hotel loans mature this year, 23% of industrial, 17% of office. Banks and credit unions hold $396 billion of it.[1] Meanwhile the CMBS delinquency rate sits around 7.35% and the special servicing rate around 11.2%, with office special servicing at 17.11%.[2]
Very little of that will be resolved by a foreclosure. It will be resolved by a modification, and the modification will be documented in a fifteen-page amendment that nobody reads as carefully as they read the original loan agreement four years ago.
What lenders are actually charging for an extension.
There is a durable story that banks facing this wall have been extending bad loans on easy terms to defer loss recognition. A Federal Reserve working paper published in May 2026 tested it against supervisory data and found the opposite: banks tightened.[3] Extensions granted after the spring 2023 stress episode were 5.1 percentage points more likely to require a principal paydown of at least 5%; for large office loans, 13.3 percentage points more likely. Spreads went up roughly 8 basis points beyond what they otherwise would have. And extended loans performed better afterward than comparable non-extended maturing loans—about 58% paid off within six quarters versus 52%.
Buried in that same list is a much smaller number that matters more than any of them. Recourse was added to previously non-recourse extensions roughly 1.6 percentage points more often than in normal times.
A paydown is priced. A spread is priced. Both get negotiated in the term sheet, both show up in the model, and both are obvious to everyone in the room. Recourse is not a price at all. It is a change in who is liable, and it typically arrives as a defined-term edit in a document the principal signs personally, at the end of a process everyone is relieved to have finished.
Above the line, below the line.
Every non-recourse CRE loan has a carve-out list, and the list is boring enough that it reads like housekeeping: pay the taxes, keep the insurance in force, deliver the financials, don’t commit waste, don’t misapply insurance proceeds, don’t file a voluntary bankruptcy, don’t transfer the property, don’t incur other debt, maintain the single-purpose entity separateness covenants.
The list is not the risk. The risk is which side of a line each item sits on.
Practitioners call it above-the-line and below-the-line. An above-the-line, or loss, carve-out makes the guarantor liable for the lender’s actual damages caused by the bad act. A below-the-line, or springing recourse, carve-out makes the entire loan fully recourse—principal, accrued interest, fees, costs—the moment the trigger fires.[4]
The same words describe both. The consequence differs by two orders of magnitude.
In ING Real Estate Finance v. Park Avenue Hotel, the borrower paid $278,759 of property taxes nineteen days late. The lender read the carve-out as springing full recourse on a $145 million loan. The court refused, reasoning that on that reading one day of tax delinquency could produce up to $90 million of personal liability.[5] That is the good outcome, and it required litigation, and it turned on a judge’s willingness to call a reading unreasonable rather than on anything the parties had written down.
The bad outcomes are better documented. In Wells Fargo Bank v. Cherryland Mall, the loan documents contained a covenant that the borrower remain solvent. The property failed the way properties fail; the borrower became insolvent; the Michigan Court of Appeals held in 2011 that insolvency breached the covenant, that the breach tripped the carve-out, and that the guarantor owed the $2.1 million deficiency.[6] Courts have similarly enforced insolvency triggers in 51382 Gratiot Avenue Holdings, and full recourse for an unpermitted second mortgage in CSFB 2001-CP-4 Princeton Park.[5]
Cherryland is worth sitting with, because of what happened next. Michigan’s legislature passed the Nonrecourse Mortgage Loan Act with bipartisan support inside three months. It took effect on March 29, 2012, applied retroactively, and was later upheld against constitutional challenge. The statute bars post-closing solvency covenants from serving as non-recourse carve-outs at all, on the stated ground that using one that way “is inconsistent with the nature of a nonrecourse loan,” is “an unfair and deceptive business practice,” and is “against public policy."[7]
A state legislature does not normally void a freely negotiated commercial term. It did here because the term was doing something neither side believed it was doing. And Michigan is one state. In the other forty-nine, the solvency carve-out is still a solvency carve-out.
The trigger is keyed to the moment you cannot pay it.
This is the structural defect, and it is not an accident of drafting. Insolvency, SPE-covenant breach, failure to pay taxes, failure to fund insurance, a mechanic’s lien the borrower disputes and never agreed to—every one of these becomes more likely as a property deteriorates. The carve-outs a sponsor most needs to survive are precisely the ones written to fire when the sponsor is in trouble.
A carve-out aimed at bad faith—fraud, misapplied rents, waste—makes sense as a deterrent, because the guarantor controls whether it happens. A carve-out that fires on insolvency deters nothing. It just converts a market outcome into a personal one.
Three things an extension does to a guaranty that nobody prices.
1. The rescue capital can be the trigger. Funding a 5% paydown usually means new money: mezzanine debt, preferred equity, a capital call with a new partner, a pledge of equity interests. Carve-out lists routinely prohibit unpermitted “indebtedness” above a dollar threshold and any “transfer or encumbrance,” direct or indirect. The financing that saves the deal can trip the provision that makes the sponsor personally liable for it.
2. Reaffirmation is a new signature, not a formality. Extension documents almost always include a guarantor consent and reaffirmation. That is a fresh execution of the carve-out obligations, and it is where a lender can widen the list, add an estoppel, waive defenses, or restart a limitations period—in a document captioned as consent to something the guarantor already agreed to.
3. “Adding recourse” does not have to mean full recourse. A lender does not need to convert the loan. Moving one existing item from above the line to below it accomplishes the same thing and edits far less text. The term sheet still says non-recourse.
Why this is a portfolio question, not a document question
A sponsor with forty loans does not have one guaranty. They have forty, written by six lenders across three vintages and a dozen states, on forms that drifted between 2014 and 2022. Some carve-outs are loss-only throughout. Some put insolvency below the line. Some are governed by Michigan law and some are not. Some have already been reaffirmed once.
The question that matters this year is not “what does my guaranty say.” It is “which of my forty guaranties turn a workout into personal liability, and which of those are maturing in the next four quarters.” That is a distribution, not a document, and it is the same shape as every other clause-level exposure: the operating expense ratchet that only travels upward, the power clause in a data center lease that decides who absorbs a utility increase, the remedies provision that determines whether a buyer gets the land or a check. Silence selects a default. So does a line break in a carve-out list.
Reading forty guaranties before you sign the fortieth extension#
Answering which of our loans put insolvency below the line, and which of those mature before June? by pulling forty loan agreements, their guaranties, and every prior amendment is exactly the work that used to consume associate weeks—and the answer expires the moment the next modification is signed.
This is the problem Acrebase is built for. Acrebase is AI-powered contract intelligence for commercial real estate: it parses loan documents, guaranties, leases, and amendments into structured data, flags unusual or missing provisions, and lets you query an entire contract library in plain language. Ask which guaranties are loss-only, which carry springing full recourse, which include a post-closing solvency covenant, which prohibit indirect transfers of equity interests, and which have carve-outs that a pending mezzanine facility would trip—and get answers with every extraction traced back to the exact language in the source document, because a liability you cannot cite is a liability you will discover in a demand letter.
For teams who would rather catch it in the amendment than in the deficiency judgment, the same review sits inside Microsoft Word. The Acrebase Word add-in puts chat, proofreading, a clause library, and redlining in the task pane next to the document, so a reaffirmation that quietly widens the carve-out list gets flagged while it is still a redline. It is the working version of the argument we made about AI in CRE due diligence: software handles extraction and comparison across the portfolio, the attorney keeps the judgment. And because extension negotiations run long and finish late, it is worth remembering what long negotiations do to deals—the guarantor consent is signed at the end, by the person with the most to lose, on the day everyone wants it over.
The practical takeaway: before you sign an extension, find three things in the guaranty—the insolvency provision, the definition of permitted indebtedness and transfers, and the exact boundary between loss liability and full recourse. Then read the reaffirmation as a new guaranty, because that is what it is. The maturity wall is not primarily a story about defaults. It is a story about several thousand sponsors accepting a small amount of recourse in exchange for time, one amendment at a time, and not all of them are counting.
Footnotes
[1] Mortgage Bankers Association, “17 Percent of Commercial and Multifamily Mortgage Balances to Mature in 2026” (Feb. 9, 2026) (2025 Commercial Real Estate Survey of Loan Maturity Volumes). https://www.mba.org/news-and-research/newsroom/news/2026/02/09/17-percent-of-commercial-and-multifamily-mortgage-balances-to-mature-in-2026
[2] Trepp CMBS delinquency and special servicing rates, 2026 monthly series, as compiled in “2026 CMBS Delinquency Rates,” Multi-Housing News. https://www.multihousingnews.com/cmbs-delinquency-rates/
[3] David Glancy, “Pretend or Amend? On Evergreening in CRE,” Finance and Economics Discussion Series 2026-025, Board of Governors of the Federal Reserve System (May 2026). https://www.federalreserve.gov/econres/feds/pretend-or-amend-on-evergreening-in-cre.htm
[4] “Non-Recourse Carve-Outs: Borrower and Guarantor Considerations,” ArentFox Schiff (loss-based carve-outs versus full recourse; insolvency and separateness triggers). https://www.afslaw.com/perspectives/alerts/non-recourse-carve-outs-borrower-and-guarantor-considerations
[5] “Your Non-Recourse Loan Isn’t Non-Recourse: Read the Bad-Boy Carveout Before You Sign,” Murphy PC (discussing ING Real Estate Finance v. Park Avenue Hotel, 51382 Gratiot Avenue Holdings, and CSFB 2001-CP-4 Princeton Park). https://murphypc.com/news/your-non-recourse-loan-isnt-non-recourse-read-the-bad-boy-carveout-before-you-sign/
[6] Wells Fargo Bank, N.A. v. Cherryland Mall Ltd. Partnership, Michigan Court of Appeals (Dec. 27, 2011). https://www.casemine.com/judgement/us/5914f99eadd7b049349a3669
[7] “Michigan Nonrecourse Mortgage Loan Act Held Constitutional,” Dykema (NMLA effective March 29, 2012; retroactive application; constitutional challenge rejected). https://www.dykema.com/news-insights/michigan-nonrecourse-mortgage-loan-act-held-constitutional.html
Acrebase is AI-powered contract intelligence for commercial real estate — clause extraction, risk flagging, and portfolio-wide search, plus an AI contract review add-in for Microsoft Word. Learn more at acrebase.com, install the Word add-in from Microsoft AppSource, or get in touch about pricing.