The most telling figure in the 2026 commercial real estate debt tally is not the roughly $6.3 trillion headline. It is the gap in growth between holders. The lenders that keep loans on their own balance sheets, banks and insurers, expanded their CRE books by about 3% over the year to Q2. The two channels that pass risk to someone else grew more than twice as fast. Fannie Mae and Freddie Mac multifamily lending rose 8.2%, and securitized CMBS balances rose 7.4%. With the 10-year Treasury at 5.28% and the curve positively sloped, this is not a market that has cleared its rate problem. It is a market where the cost of that problem is being moved off bank equity and onto federal guarantees and yield-hungry bond buyers.
Wolf Richter's framing in WOLF STREET, "water under the bridge," captures the attitude. The more useful observation is structural. The private balance sheets that took the 2022 to 2025 losses are lending cautiously. The conduits built to distribute risk are lending aggressively.
Who Holds the $5.1 Trillion
The Mortgage Bankers Association splits the market into two pools. Income-producing properties, meaning completed buildings with tenants, carry $5.1 trillion of debt. A second pool of about $1.41 trillion covers properties without rental income: $696 billion of construction loans, which are taken out when the finished building is refinanced, and $712 billion of loans on owner-occupied commercial property, where the business occupying the building services the debt. The two pools sum slightly above the $6.3 trillion headline, so the breakdown is better read as proportions than as an exact reconciliation.
Within the income-producing $5.1 trillion, ownership breaks down as follows:
- Banks and thrifts: $1.92 trillion, 37.5%
- Fannie Mae and Freddie Mac: $1.17 trillion, 22.8%
- Insurance companies: $815 billion, 15.9%
- CMBS, CLOs and other securitized: $771 billion, 15.1%
- Other government: $220 billion, 4.4%
Banks also hold $454 billion of construction loans and the $712 billion of owner-occupied loans, bringing their total to a record $3.1 trillion, close to half of all CRE debt. That concentration is why bank exposure dominated the 2023 and 2024 headlines, from Aozora's disclosure of large losses on its $1.9 billion US office book in February 2024 to Deutsche Bank's repeated increases in US CRE loss provisions.
The figure the headlines skip is federal exposure. Combining the GSEs with other government holders gives about $1.39 trillion, or 27% of income-producing CRE debt. In multifamily, which accounts for $2.34 trillion, or 46% of income-producing debt, the GSEs hold half. Washington is the largest single underwriter of American apartment buildings, and its book is the one growing fastest.
A Curve That Offers No Exit
The original damage came from two rate channels that hit together. Floating-rate loans from the pandemic era reset off short rates that jumped from mid-2022. Fixed-rate loans hit a refinancing wall, because a building that worked at a 3% coupon did not work at 7%. Both channels depended on the shape of the Treasury curve.
As of October 7, the 2-year sits at 4.77% and the 10-year at 5.28%, a +0.51 percentage point spread. For CRE borrowers, both ends of the curve are now hostile. Floating-rate debt still prices off a short end near 4.8%. Fixed-rate refinancing now prices off a long end above 5%, higher than when the refinancing wall first became a crisis. A curve that has un-inverted by having long yields rise, instead of short rates falling, removes the escape route of term financing. Borrowers who extended and waited for cheaper long money have watched it become more expensive.
The transmission matters for who lends. Bank CRE lending is largely floating or short-fixed and funded by deposits, so banks feel the short end in funding costs and the long end in collateral values. GSE multifamily loans and CMBS conduit loans are usually long-term fixed-rate paper priced as a spread over Treasuries. When the 10-year is above 5%, those coupons begin from a base that only a lower purchase price can absorb. The 8.2% and 7.4% growth rates therefore imply one of two things. Either new loans are being written against substantially reset valuations, or the conduits are accepting thinner debt-service coverage than the coupon math would suggest. Which one applies is not visible in the aggregate data, and it is the most important unknown in this market.
Two Conduits, One Direction
GSE and CMBS lending share a mechanism. In both cases, the entity that sources and prices the loan is not the one carrying the full loss if it fails.
For the GSEs, multifamily credit risk is backstopped by a federal government that has controlled both companies since 2008. Some of it is laid off to private investors through risk-sharing structures, but the final loss absorber is public. A government lender growing its multifamily book 8.2% a year while long yields exceed 5% is acting countercyclically, whether or not that is the stated intent. It keeps apartment values supported by supplying capital that private balance sheets are rationing.
For CMBS, the originator earns a fee and sells the bonds, and the investor takes the credit. Richter states his suspicion directly: before the pandemic, banks securitized their riskiest office loans and let CMBS investors absorb the subsequent losses. That claim is his inference and has not been demonstrated loan by loan. The outcome it describes is documented, though. Office CMBS delinquency rose above its Financial Crisis peak, and office towers that sold after 2022 went for 50% to 80% below pre-pandemic prices. The investors who owned that paper took losses that bank equity did not.
So the same channel that distributed the last round of losses is now the second-fastest-growing holder, at $771 billion. New buyers are reaching for spread over a 10-year that already yields more than 5%. In a high-rate regime, a high absolute yield attracts buyers who might demand more structural protection if the base rate were 2%. That is a plausible driver of demand, but no investor-flow data is available here to confirm it.
The countercyclical government lender and the yield-chasing private buyer arrive at the same place from opposite motives. Both are adding CRE exposure faster than the banks and insurers that just lived through the losses.
The Case for Clean Vintages
The strongest objection is that this time the collateral is different, and it deserves full weight.
The 2019 to 2021 loans failed because they were sized off inflated appraisals at 3% coupons, often through cash-out refinancings that left little equity in place. Loans written in 2025 and 2026 carry coupons set in a regime with a 10-year near 5%, and many sit on properties bought at heavily reduced prices. A buyer who acquired an office tower at a 60% discount, financed at a current rate, has a cost basis the building's rent can actually service. That is how a credit cycle is supposed to clear: losses are recognized, assets change hands at realistic prices, and new lending starts from a sound base. On this view, 7.4% CMBS growth signals healing, not repeated error.
The bank data supports part of that argument. Banks kept their CRE books flat through 2024 and the first half of 2025, offloading troubled loans about as fast as they wrote new ones, before growing again late in 2025. Their $88 billion, or 2.9%, year-over-year increase in Q2 2026, per the Federal Reserve's Z.1 accounts, reflects lenders that have cleaned house and believe they can price risk again. Insurers, whose long horizons make them sensitive to duration mismatch, grew 3.3% to $815 billion. Neither pattern looks like panic.
The objection has two weak points. First, the GSE book is mostly multifamily, not office, and multifamily never went through the forced repricing that office did. Office sold at 50% to 80% discounts. Apartments generally did not, which means the multifamily cost basis has been reset less than the office cost basis, while multifamily lending is growing the most. Second, a clean vintage only protects investors if rates do not rise further. The 51 basis point positive slope came from long-end yields moving up. If that bear steepening continues, today's "realistic" valuations will look like the 2021 appraisals did in hindsight: correct for the rate at underwriting and wrong for the rate at maturity.
What Would Change the Read
The decision variable is the growth gap between risk-retaining lenders and risk-distributing conduits. Today it runs roughly 3% against 7% to 8%. While that gap persists with the 10-year above 5%, the reasonable conclusion is that CRE credit is being rebuilt through the channels least exposed to the originator's own losses, and that the eventual cost of any further rate shock will fall more on federal guarantees and CMBS holders than on bank capital.
Three observable conditions would weaken this view. Banks accelerating their CRE growth toward the conduits' pace would indicate that private balance-sheet lenders share the conduits' confidence. A sustained decline in the 10-year that flattens the curve from the long end would reopen term refinancing and reduce the coupon pressure on the 2025 and 2026 vintages. The deciding evidence will be delinquency data on GSE multifamily loans and post-2024 conduit CMBS once those loans have seasoned for two to three years. Clean performance there would confirm that the reset was real.
Without that, "water under the bridge" describes who took the last losses, not who is lined up to take the next ones.
