Your bank stopped lending to businesses. It lends to lenders now.

I switched banks. Last spring, in a small office, with a man who kept apologising for his printer.

He talked me through the deposit insurance. He said the bank lends my money to businesses in the area. I nodded along. Boring is what I want from a bank.

So this week I looked at what banks do with the money now.

Lending to companies is no longer the main event. The fastest-growing loan book in American banking is loans to other lenders. The bakery and the trucking firm still get the money. It reaches them through a fund now.

Regulators call those borrowers nondepository financial institutions. Private credit funds. Private equity funds. Mortgage companies. Finance companies. Broker-dealers.

In early 2010, US banks had $56 billion lent out to them. By the third quarter of last year they had $1.32 trillion. That is one dollar in every ten that American banks lend.

Fifteen years ago it was less than one in a hundred.

The bank still funds the trucking firm. It writes the cheque to a fund instead. The fund picks the borrower, sets the terms, and decides who is good for the money.

That arrangement interests me for one reason. The credit risk did not leave. It moved one step down the chain, into a company the bank does not run and the bank regulator does not supervise.

When a loan book grows more than twenty times over in fifteen years, I want to know who checked the borrowers. When the same banks then buy protection from those same funds, I want to know what the protection is worth.

So this week I followed the money out of the branch. Then I followed it back.

This connects directly to how banks describe their own risk.

When a bank lends to a nonbank lender, the loan is usually secured, and the advance rate against the collateral is conservative. The bank looks at the fund. It does not look at the fund's borrowers.

On the banks' own books, those loans almost never go bad. In the third quarter of last year, 0.15 percent of them were past due or on nonaccrual.

Loans to ordinary companies were late at 1.32 percent. Close to nine times the rate.

The pattern holds at every size of bank. At the biggest banks the figure is 0.12 percent. At the smallest, 0.49 percent.

I have no idea whether it holds. Most of these funds have never run a portfolio through a full credit downturn, and the collateral behind the bank's loan is loans the fund wrote itself.

In December the rules loosened. On the fifth, the FDIC and the OCC withdrew the leveraged lending guidance they had issued in 2013, saying it had become too restrictive.

And here's where it spreads.

Most of what banks give these lenders is a revolving line rather than a term loan. The money sits available and undrawn until the fund wants it.

At the end of the third quarter of last year, undrawn commitments to nonbank lenders came to $987 billion. That is 42.9 percent of everything banks have promised them.

A fund draws its line when its other funding stops. Which is the same week the bank would rather hold on to its cash.

The exposure sits in very few places. Banks with more than $100 billion in assets hold about 86 percent of the total. Ten institutions hold roughly 71 percent.

At the end of 2024, four banks held 47.8 percent of the industry's whole exposure between them: JPMorgan, Bank of America, Wells Fargo and Citibank. So this does not sit spread across four thousand banks. It sits at the handful everything else leans on.

Meanwhile, the traffic runs the other way too.

Banks have spent a decade selling the credit risk on their loan books to outside investors. The instrument is called a synthetic risk transfer. The bank keeps the loan and keeps the customer. It buys protection against the losses.

The Bank for International Settlements sized the market in March. Annual issuance went from under €5 billion in 2016 to €21 billion in 2024. Outstanding protection now covers close to €800 billion of loans.

Most of the buyers are credit funds, hedge funds and asset managers. The same sort of firms banks lend to. European regulators have a term for what follows, and the BIS borrowed it: circles of risk.

The risk leaves one bank, goes into a fund, and comes back through a different bank.

That second bank financed the fund's purchase. The BIS also describes European and Japanese banks passing credit risk to American hedge funds and credit funds while lending to some of the same firms.

The relief is modest so far. About 43 basis points of core capital for the average issuing bank, against sector capital ratios of 14 to 16 percent.

But the BIS made one point clearly. This market has not yet lived through a long credit downturn. The protection has never been tested on a bad day.

Three more things worth keeping track of.

Wholesale electricity across the PJM grid cost $136.53 per megawatt hour in the first quarter, against $77.78 a year earlier. That is a 76 percent jump, the largest one-year rise in the grid's history, and PJM serves about 65 million people across thirteen states and Washington DC. Monitoring Analytics, the market's independent monitor, puts data centre demand behind 63 percent of the capacity increase, worth roughly $9.3 billion that customers pay through their bills.

The Korean won's inflation-adjusted value fell to 82.99 in June on the BIS index, the weakest since March 2009. Of the 64 economies the BIS tracks, only Japan sat lower, at 65.3. Then it turned. Through July 24 the won gained 6.24 percent against the dollar, the best of the twenty major currencies Yonhap Infomax follows. A currency that cheap that moves that fast tends to pull other things with it.

Private equity funds are not sending cash back. MSCI puts distributions at roughly 6 percent of buyout assets in the year to June 2025, against a ten-year average near 14 percent. McKinsey's 2026 private markets report finds five-year rolling distributions for buyout funds at the lowest level it has recorded. Pension funds and university endowments are the ones waiting. We'll see.

In 1988, a small team at Citibank in London built something new. They called it Alpha Finance Corp.

It borrowed short and lent long. It sold commercial paper a few months at a time and bought high-grade bonds with the proceeds. The spread was the profit.

The clever part was where it sat. It sat off the bank's balance sheet. The bank managed it, earned fees from it, and owed it nothing.

The design got a name. A structured investment vehicle. By the summer of 2007 the industry ran around $400 billion through them. Citigroup advised seven.

In its filing for the third quarter of 2007, Citigroup told investors it had committed $10 billion of liquidity to those vehicles, of which $7.6 billion had been drawn. In the same filing it wrote that it would not take actions requiring it to consolidate them.

The bank owed those vehicles nothing. It had put that in writing, six weeks before it paid.

On December 13, 2007, Citigroup took all seven onto its balance sheet. Forty-nine billion dollars of assets, and the debt that funded them.

The assets were not the trouble. By the bank's own account at the time, the portfolio was highly rated and carried no direct exposure to subprime mortgages.

The funding was the trouble. The short-term paper market stopped rolling over. Moody's and S&P put the vehicles' senior debt on review for downgrade.

And a bank cannot let something carrying its name fail, whatever the paperwork says. Reputation is not a legal obligation. It behaves like one anyway.

That is what I keep coming back to. The risk sat legally somewhere else for nineteen years. It came home in one evening.

Contracts describe where risk sits. They do not decide where it goes.

We'll see.