Everyone knows banks pulled back from risky real estate lending after 2008. What most people don't know is that they never actually stopped. They just moved up one layer in the capital stack and started lending to the debt funds doing the risky lending instead. That distinction is now worth $1.57 trillion, has been growing at 22% annually for 15 years, and sits directly in front of the worst commercial real estate credit cycle in a generation.
Loans to non-depository financial institutions (NDFIs), the funds, mortgage companies, and private credit vehicles that do the lending banks no longer do directly, now represent 52% of the entire U.S. banking industry's Tier I capital and allowance for credit losses. At banks with over $100 billion in assets, that figure is even higher at 68%. The largest institutions in the country have NDFI exposure equal to more than two-thirds of the capital buffer regulators require them to hold against losses, according to the FDIC's February 2026 report.
On the surface, everything is fine. NDFI loan delinquency sits at 0.14%. Bank executives describe their portfolios as "stagnant." Analysts call the credit quality "largely a nonissue." What sits inside that $1.57 trillion number tells a different story.
Fifteen Years of Compounding
Regulators began tracking bank lending to NDFIs in Q1 2010 during the aftermath of GFC-era NDFI failures. Then the total was $56 billion. By Q4 2025, per S&P Global Market Intelligence, it stood at $1.57 trillion, reflecting a compound annual growth rate of 21.9% over 15 years. The next fastest growing loan segment during the same period was multifamily commercial real estate, at 7.7%. NDFI lending grew at nearly three times that pace, consistently, through rate cycles, recessions, and periods of market stress.
NDFI lending is how banks maintain real estate exposure without appearing to. The loans do not show up as CRE on their books but instead as credit facilities to financial intermediaries. The risk is the same, but the classification is different. In Q4 2025 alone, NDFI loans grew $129.7 billion, while every other loan category combined grew $31.4 billion. JPMorgan Chase holds $237.9 billion in NDFI exposure. Wells Fargo holds $212.1 billion. Ten banks control 71% of the total, and seven of the top twenty grew their books by more than 10% in a single quarter.
Although this is the fastest growing bank asset category in U.S. banking, it is also the one that looks the least like what it actually is.
How the Machine Works
Banks' retreat from risky lending after 2008 did not shrink the market for risky loans. Instead, it created it. Private credit now represents 30% of all below-investment-grade corporate debt, up from 13% post-GFC. The capital banks stopped deploying found a new home in lightly regulated fund structures that then borrowed from those same banks to operate.
Every capable CRE debt fund operates on back-leverage. The fund raises LP equity, deploys it into CRE loans at 9-11% yields, then pledges that loan portfolio to a bank and borrows against it at a cheaper rate. The spread between what the bank charges the fund and what the fund charges the borrower is where the return premium comes from. Without the bank line, the economics do not work.
Since 2020, more than $137 billion has been raised by closed-end CRE debt funds. Private credit AUM grew from $2 trillion to $3 trillion between 2020 and 2025. At the 2021 peak, the sector lent over $900 billion into the CRE market, mostly floating-rate bridge loans on transitional properties built for a zero-rate world.
Then the Fed raised rates 525 basis points in 18 months.
When a bank extends a warehouse line to a CRE debt fund, it underwrites against the fund's aggregate borrowing base, not the individual loans underneath. The bank sees a pool of collateral with advance rates applied. It cannot see in real time whether underlying borrowers are current, deferring payments, or in quiet default. The FDIC put it plainly: "For loans originated by NDFIs, the credit decisions, credit administration, and collateral valuations fall outside the direct control of banks."
In a stress scenario where multiple funds sell simultaneously, the collateral value is set by whatever the last distressed trade cleared. When one major publicly traded mortgage REIT bulk-sold $494 million in bridge loans at 74 cents on the dollar, they weren't the only ones hurt. It told every bank holding a warehouse line against similar assets what their collateral is now worth.
The Vintage That Cannot Refinance
Loans originated at the 2021 peak were written at 3.5-4.5% all-in rates. Those loans now carry 7.5-8.5% or higher, a 90% increase in debt burden on properties whose net operating income never came close to doubling.
Consider what that means for a sponsor trying to hold on. Interest rate cap renewals have repriced to consume roughly 13% of NOI. A borrower in that position is not managing a stressed loan. They are managing an insolvent one.
The LTV math is equally brutal. CRE values are down approximately 20% from their 2022 peak. A loan originated at 70% LTV on a $100 million property now worth $80 million sits at 87.5% effective LTV. New lenders offer 55-60% LTV, creating an equity gap of $26-31 million the original sponsor cannot fund. Every aggressive resolution strategy, deed-in-lieu, note sales, borrower bankruptcy, is accelerating. The one that is not accelerating is loan modification and extension, which remains the largest category by volume but is running out of runway as borrowers exhaust their options.
Extend and Pretend
50% of all CRE CLO loans are past maturity and non-performing, per CRED iQ data from mid-2025. Only 14.3% of the pool is current. Morningstar DBRS projects that over 50% of the $100 billion in securitized loans maturing in 2026 will not pay off at maturity. The CLO distress rate grew from 1.4% in early 2023 to 14.4% by January 2026, a ten-fold increase in 26 months. For context, that is roughly double the peak delinquency levels during the 2008 financial crisis.
Per KBRA, 27% of private credit borrowers had interest coverage ratios below 1.0x in 2024. More than one in four borrowers could not cover their interest expense with operating income in an economy with no recession.
Those losses have not been realized. Private debt funds operate as closed-end LP structures with no mark-to-market requirement. Losses are recognized on resolution only. When a borrower cannot pay, the fund modifies the loan, extends the maturity, and allows unpaid interest to roll back into the loan balance. This is called PIK, payment in kind. The debt grows while the lender books income it has never received, the loan stays on the books at or near par, and the delinquency rate stays clean. Nothing is resolved. The losses just move forward in time.
The publicly traded mortgage REITs show you what the private funds will not. One publicly traded mortgage REIT has bulk-sold hundreds of millions in bridge loans at 74-83 cents on the dollar, reported distributable losses, cut its dividend 50%, and now trades at roughly 0.20x book value. The market is pricing near-total equity impairment. That is a fund that marks to market and must disclose. The private debt funds sitting behind bank warehouse lines are neither of those things, and their portfolios are carrying the same vintage of loans.
Where the Risk Actually Lands
The distribution of this exposure is not uniform, and the banks most at risk are not the ones people are watching. JPMorgan and Wells Fargo dominate the headline NDFI figures, but their portfolios are diversified across all five subcategories. Banks with $10 to $100 billion in assets have nearly 40% of their NDFI portfolios concentrated in mortgage credit intermediaries, the most CRE-exposed category. These institutions carry less capital, less diversification, and a more direct line to the CRE market that is working through its worst credit cycle in decades. A stress event in CRE debt funds hits the regional bank tier harder on a relative basis than it hits any too-big-to-fail bank.
Add the unfunded exposure. In Q3 2025, unfunded commitments to NDFIs were $987 billion, 42.9% of total NDFI commitments. These are revolving credit lines funds can draw on whenever they need liquidity. In a stress scenario, every fund reaches for that capacity at the same time, at the exact moment the collateral backing those facilities is deteriorating. The FDIC described the result: "A fire sale might decrease bank asset values at the same time that NDFIs may be having trouble repaying bank loans."
None of this stays inside the fund structures. When funds are forced to sell, they set prices that require banks to write up their loss reserves, which shrinks book values and triggers more selling. The cycle is self-reinforcing. When the funds start losing, those losses have to land somewhere, and they land on the banks.
The fastest growing bank asset category for 15 years sits in front of a 50% past-maturity loan pool, a distress rate that went from 1.4% to 14.4% in 26 months, and a maturity wall that more than half of 2026's securitized maturities will not clear.
The Last Guardrail
This is the structure that was built. Banks stopped lending directly to risky borrowers. Debt funds filled the gap because sponsors needed leverage banks would not provide. Banks then financed the debt funds, transferring the risk into the shadow system. The funding stayed in a regulated loop that has been running for 15 years.
On December 5, 2025, the FDIC and OCC rescinded the 2013 Interagency Guidance on Leveraged Lending, the last formal constraint on how much leverage bank-funded borrowers could carry, while the CLO distress rate sat at 14.4% and rising.
The structure built to defer risk is meeting a maturity wall it cannot extend past. The capital behind it is running out of options. And the last guardrail just came down.
Sources: FDIC "Banking Issues in Focus: Bank Lending to Nondepository Financial Institutions," February 2026 (Fritzdixon & Lu); S&P Global Market Intelligence, "US Banks' NDFI Lending Pace Reaccelerates in Q4 2025," February 26, 2026 (Lawson & Villaluz); CRED iQ CRE CLO Distress Monitor, 2025-2026; Morningstar DBRS; KBRA Private Credit Default Monitor, 2024
