Two Federal Reserve blog posts, published a month apart, tell you everything you need to know about why the official reassurance on shadow-bank stress is aimed at the wrong target.
The first, from May 2026, reports that NBFI troubles "do result in 'stress and strain' for banks," while relaying the consensus that this is "unlikely to amount to systemic concerns" and that policymakers are "watching carefully."
The second, from April 2026, surveys 160 years of American banking history and concludes that, per the Fed's historians, insolvency is usually a necessary condition for failure and that runs, while an important trigger, mostly accelerate failure at already-weak banks rather than cause sound banks to fail. They fail from insolvency—assets that turn out to be worth less than the balance sheet claims.
Put those two findings side by side and a contradiction surfaces. "Watching carefully" is watching for the wrong signal.
Start with what already happened, because it is more concrete than any model.
JPMorgan, Fifth Third, and Barclays wrote warehouse loans to Tricolor, a subprime auto lender. These were senior claims—the safe part of the capital structure, the part that gets paid first.
They still charged off roughly half a billion dollars combined: JPMorgan a $170 million charge-off, Fifth Third a $178 million impairment, Barclays a £110 million hit, about $148 million.
The reason is the load-bearing detail of this entire episode. Tricolor allegedly pledged the same auto loans to multiple lenders as collateral, duplicating vehicle identification numbers to draw multiple loans against a single car.
So the collateral behind the "senior" claim was not there. A senior loan against an asset that has been sold twice is not senior to anything. It is unsecured, dressed as secured.
A senior loan against an asset that has been sold twice is not senior to anything. It is unsecured, dressed as secured.
This is the core question the piece turns on: if senior claims can vaporize, what does "bank exposure" actually mean?
The standard answer treats exposure as a liquidity problem. The worry, implicitly, is that an NBFI draws down its credit lines all at once, or that redemptions cascade, and a bank gets caught short of cash. That is the run framing.
It is also why the reassurance sounds soothing. If the danger is a run, the toolkit is well understood—deposit insurance, the discount window, a lender of last resort. Liquidity events are survivable because central banks can manufacture liquidity.
But the Tricolor loss was not a liquidity event. No one ran. The banks simply discovered that an asset on their books was worth a fraction of its stated value, and they wrote it down.
That is insolvency-first, in miniature. And the Fed's own historians say that is how banks actually die.
Failing banks consistently show declining income, eroding capital, rising asset losses, and heavier reliance on expensive funding in the years before they go under—with and without runs, with and without deposit insurance.
The recovery-rate evidence seals it. Before federal deposit insurance arrived in 1934, unsecured depositors of failed banks recovered only 66 cents on the dollar.
If those banks had been solvent institutions killed by panic, their assets would have covered their debts once the dust settled. They did not. The assets simply weren't worth enough. Per the Fed's historians, most such banks were already insolvent when the run arrived; the run largely set the time of death.
Examiners agreed. In the pre-insurance era they classified only 36 percent of failed-bank assets as "good," 47 percent as "doubtful," and 18 percent as "worthless." The recorded causes of death were poor local conditions, asset losses, and fraud.
Now hold that word—fraud—against the two 2025 bankruptcies.
First Brands, the auto-parts supplier, financed itself through an opaque web of off-balance-sheet vehicles secured by receivables. Traditional loan covenants failed to detect the hidden commitments.
The result: lenders who believed they were underwriting debt at roughly 5x leverage were actually closer to 20x. Court filings say about $2 billion is unaccounted for. As with Tricolor, the same collateral may have been pledged to multiple lenders.
Fraud-driven double-pledging produces precisely the failure mechanism the 160-year study identifies: an asset that isn't worth what the balance sheet says. Not a funding gap—a valuation lie that only surfaces on the way down.
This is why the "senior" label is so dangerous here. Seniority is a claim on the recovery waterfall. It assumes the collateral exists and belongs to you. Fraud attacks that assumption directly, and when it does, seniority is a legal fiction.
The Tricolor noteholders learned this fast. Some notes are now trading at under 10 cents on the dollar, and investors are suing JPMorgan, Barclays, and Fifth Third, alleging 2022 and 2024 audits flagged that nearly half of daily borrower payments were routed to wrong accounts and that recoveries were booked on repossessions that never happened. They call it a "Ponzi-like fraud."
That is the transmission channel, and it moves solvency losses, not liquidity shocks, from the shadow-bank sector into regulated banks.
Now the scale. As of year-end 2025, about fifty bank holding companies had total credit exposures to NBFI borrowers—on-balance-sheet loans plus undrawn commitments—exceeding 100 percent of their Tier 1 capital. The most extreme sat at 4 to 6 times Tier 1.
Many of these are regional banks, with $10–100 billion in assets. Bank lending to NBFIs, the Fed notes, accounted for essentially all of bank lending growth in 2025.
And the market has already priced the linkage. Around three separate NBFI stress events, banks with higher NBFI exposure showed negative abnormal returns even after stripping out market-wide moves. The stocks knew the channel was real before the reassurances arrived.
Here is where the story stops being a modeling debate and becomes a policy failure.
If insolvency is the killer, the antidote is not liquidity backstops. It is capital, supervision, and risk management—the things that catch a mispriced asset before it detonates.
Which is exactly what is being cut. According to reporting, on October 30, 2025, the Fed would reduce bank supervision staff by 30 percent.
In December, Senate Banking members wrote to the Fed, OCC, and FDIC warning that large banks had already reported "more than a billion dollars of losses on just a handful of poorly underwritten commercial loans"—and called it "likely just the tip of the iceberg."
Their charge is direct: the agencies are cutting examiners, limiting examiner authority, and letting banks deplete cushions "at the exact moment you should be strengthening it."
Stack the pieces. The real failure mechanism is solvency erosion from bad assets. The exposure is concentrated at regional banks with thinner cushions and less market scrutiny. And the workforce whose job is to find bad assets before they metastasize is shrinking by nearly a third.
There is a live illustration of how fast these portfolios can turn. In late January 2026, Anthropic's launch of AI automation tools intensified concerns about the credit quality of software-sector loans—a reminder that a technology shock can become a credit-quality shock inside NBFI books overnight.
The point is not that a crisis is imminent. Default rates remain below recessionary highs, and the individual failures may well prove idiosyncratic.
The point is narrower and harder to dismiss. The reassurance that this is "non-systemic" is not a hedge; it is a category error.
It answers a question about runs—will depositors panic, will liquidity freeze—when the mechanism actually threatening these banks is the quiet one: senior loans discovered to be junior to fraud, written down one impairment at a time.
Runs announce themselves. Insolvency accumulates silently, on balance sheets, in collateral pledged twice, in leverage that was 20x and looked like 5x.
Watching carefully for a run means staring at the door while the foundation rots. The Fed's own historians already wrote down which one kills the bank.