It is bank earnings week, and people are worried about private credit. You could imagine banks gloating. “People are worried about the credit quality of private-credit loans,” they might say, “and are rushing to take money out of retail private credit funds. We, meanwhile, never got caught up in the private credit hype cycle, and we have a fortress balance sheet and access to lots of cheap funding. This is a great opportunity for us: As private credit cools down and retreats from making aggressive loans, pricing will get more favorable and we will gain market share.”

That is: One model of private credit is that it is a source of credit that competes with banks, and bad news for private credit is good news for its competitor, banks.

In fact, though, there are a lot of stories like this:

Wells Fargo & Co. said its exposure to private-credit firms was roughly $36.2 billion in the first quarter, offering details on a category closely watched by investors. 

Companies in the business services, software and health-care industries make up about half of the total value of the collateral, with software companies accounting for 17%.

For more than 98% of those transactions, Wells Fargo can adjust the margin if the credit performance of the underlying assets worsens. The securitized loans would have a roughly 40% cushion, meaning the funds the bank lends to would absorb about 40% of losses before they are recognized by Wells Fargo, according to an investor presentation Tuesday.

Or this:

Wall Street banks reported at least $100 billion of exposure to private-credit firms, offering a glimpse at what has become a closely watched industry with many investors on edge over credit quality and the growing impact of artificial intelligence.

Wells Fargo & Co. said its exposure to private-credit firms was roughly $36.2 billion in the first quarter, while Citigroup Inc. reported $22 billion of such loans in the fourth quarter, noting that it’s had zero losses over the life of the portfolio.

JPMorgan Chase & Co. put the figure at $50 billion, with Chief Executive Officer Jamie Dimon saying on a call with analysts that he wasn’t “particularly concerned about it.”

“You have to have very large losses in private credit before at least it looks like banks are going to get hit,” Dimon said. “It doesn’t mean you won’t feel some stress and strain and you might have to do something about it. But I’m not particularly worried about it.”

That is, the story of private credit in 2026 is not so much that private credit and banks are distinct sources of credit, but rather that “banks are re-tranching”: Private credit firms make loans to companies, funded in part with their own investors’ money and in part — 60%, in Wells Fargo’s portfolio — with money that they borrow from banks. Companies borrow money, the loans are pooled in private credit funds, and if the loans default, the private credit investors take the first 40% (give or take) of losses and banks take the next 60%.

Broadly speaking, that should make the financial system safer. Banks are highly leveraged and funded with demand deposits; if people get nervous there might be a run on the bank. Private credit funds are less leveraged (60%?), and the whole point is that their investors can’t take their money back. So there are no real runs on private credit, no real fire sales of loans, no real drying up of credit due to investor fickleness. “BlackRock Sees Private Credit Tumult as Way to Take Market Share,” Bloomberg reports today, even though many investors in BlackRock’s retail private credit fund asked for their money back this quarter. BlackRock just said no! Problem solved.

So Jamie Dimon is more or less right to be sanguine. Still, you could imagine two problems with this sort of re-tranching:

1. Sloppiness. If your model is “instead of lending $60 to businesses, banks lend $60 to private credit funds, which then lend $100 to businesses,” then this re-tranching of banking creates more loans than the banks would have made on their own. How does the system make more loans? Maybe by making worse loans: loans to riskier companies, or loans at higher leverage ratios that make companies riskier, or loans to frauds. A senior claim on worse loans might be worse than full exposure to better loans.
2. Correlation. What are the chances that all the private credit loans will default at once, leaving private credit funds with losses of more than 40% and thus causing losses at the banks? I mean, normally, quite low: These are pools of loans to lots of different companies, the economy is mostly fine, and you wouldn’t normally expect losses greater than 40% absent widespread economic devastation. Unless the loans are in some functional sense all the same: If every private credit loan is fundamentally “we loaned a lot of money to finance a leveraged buyout of an enterprise software-as-a-service company,” and if modern AI models render all of enterprise SaaS worthless, then, yes, all those private credit loans could go to zero and the banks could lose money. This seems in several ways sort of an unlikely scenario, but surprising correlations do cause financial crises, and I suppose that’s why Wells Fargo takes pains to explain that only 17% of its private credit loans are in software.

The private credit thing is an annoying market failure. What is happening is, roughly:

1. A lot of private credit funds take the form of permanent non-traded business development companies (BDCs). Investors (largely individuals) bought shares from a BDC, and the BDC used their money to make loans. The shares do not trade anywhere. The only way for investors to get their money back is that, once a quarter, the BDC offers to buy back up to 5% of the shares, at net asset value. If holders of more than 5% of the shares want to redeem, tough luck; they only get back 5%.
2. Holders of more than 5% of most of the BDCs (except one Goldman Sachs one!) have asked to redeem this quarter. 
3. That might be a matter of generic nervousness, but a simple explanation is that the investors do not trust the BDCs’ net asset values. A BDC might tell its investors “all our loans are doing great and they’re worth 100 cents on the dollar,” but investors might not believe that; they might think that some of the loans will default and they’re only really worth 80 cents on the dollar. (Other private credit funds, traded BDCs, trade at market prices that often reflect 20% discounts to NAV.) If you think that your fund is worth 80, but the fund will cash you out at 100, you should cash out at 100, so people do.
4. Meanwhile a lot of people — the BDC managers, Boaz Weinstein, some institutional investors — think that all of these concerns are overblown, that private credit is doing great, and that now is a great time to buy low. But they can’t buy low. Non-traded BDCs can generally only transact at net asset value. (And net asset values have not fallen much.) This means that (1) the BDCs themselves cannot cash out redeeming holders at 80 cents on the dollar, even if a lot of BDC investors would take that deal, and also (2) other investors can’t put money in at 80 cents on the dollar: If they want to put money in, they have to pay 100 cents on the dollar. Some of them do that! These private BDCs continue to get inflows. But the inflows are usually smaller than the outflows, because obviously the market-clearing price is something less than 100 cents on the dollar.
5. The obvious solution here is to have some sort of secondary market where people who want out of BDCs can match with people who want in at a discount. People who want a private-credit bargain can bid 80 cents on the dollar for BDC shares, and existing shareholders of that BDC who want out — and who can’t get all of their money back from the normal 5% tender — can sell to the bargain-hunters.

This is approximately what Boaz Weinstein is up to, tendering for a Blue Owl BDC at about 65 cents on the dollar. But he’s just one buyer. There is not a more general two-sided market matching buyers and sellers of discounted non-traded BDCs, because those BDCs are non-traded. That’s their whole thing. There are publicly traded BDCs where anyone can buy and anyone can sell, and those really do trade at discounts to NAV, but they’re different. The non-traded BDCs don’t trade.

But what if they did? On Friday, Bloomberg’s Natasha Voase and Claudia Cohen reported:

As a growing number of French property funds freeze withdrawals in the face of a client exodus that’s overwhelmed their cash buffers, the true cost of liquidity is becoming clearer.

After halting redemptions earlier this year, a fund managed by Praemia REIM France has begun allowing investors to offer their shares on a secondary market, matching buyers and sellers. The initial trades show the high price some shareholders are willing to pay to get their money back.

An investor in Praemia’s Priompierre vehicle swallowed a discount of 60% to the fund’s last subscription price in the initial secondary market transaction that took place in the last week of March, the company’s data show. The only buy order so far for clients attempting to extricate themselves from Perial Asset Management’s Perial O2 fund is at a 70% discount.

Praemia’s secondary market offering is in its infancy and transaction volumes are low, a spokesperson said in an emailed statement.

“The price resulting from this first session should be interpreted with caution, as it primarily reflects a one-off match between a small number of buy and sell orders, rather than the underlying value” of the portfolio, they wrote.

Obviously French property funds are not US private credit funds, but the basic idea applies. If you run a non-traded fund, and it has an official net asset value, and at that value more people want out than want in, why not let the people who want out sell — at market-clearing prices — to the people who want in? You know who wants out (they are pestering you), and you probably know who wants in (you are pestering them); you are best situated to set up your own informal two-sided secondary market to match buyers and sellers. And then, obviously, you send out a statement saying that the market-clearing price does not reflect the underlying value of the portfolio. 

You could have a model of carbon removal credits that is like:

1. Each ton of carbon that is added to the atmosphere will make humanity worse off by $X.
2. Microsoft Corp. has many human customers, employees and shareholders, and it has offices and data centers in places that will be affected by climate change. Therefore, each ton of carbon will make Microsoft worse off by $Y.
3. There are people who will remove carbon from the atmosphere if you pay them $Z per ton.
4. If $Y is greater than $Z, Microsoft should pay those people to remove carbon. 
5. Even if it isn’t, maybe there’s some sort of collective-action dynamic where Microsoft and 100 other big companies will each pay to remove some carbon, and the collective benefit to those companies from removing all that carbon will be greater than the amount they collectively pay. Of course other companies that don’t participate will also benefit, but a big important company like Microsoft can’t free-ride like that; Microsoft’s participation helps encourage others to participate and is thus valuable to Microsoft.

Something like that; Step 5 is crucial (nobody thinks the benefit per ton to Microsoft is greater than the cost per ton) but pretty fuzzy. But this model seems pretty conventional. 

You could have another model:

1. It is good public relations for a big US company to say “we are carbon neutral.”
2. Therefore, there is some benefit to Microsoft in adding up its carbon emissions and paying some people $Z per ton to remove that much carbon: Then Microsoft can say “we are carbon neutral.”
3. If the public relations benefit of saying “we are carbon neutral” is greater than $Z times the number of tons of carbon Microsoft emits, it should do this.

This model is also a little fuzzy (it’s hard to quantify the PR benefit), and I suppose it is additive to the previous model (you get the PR benefit and the substantive climate-change benefit). But if you think that this second model primarily explains how US companies think about carbon emissions, then you might make the following predictions:

1. As Microsoft’s own carbon emissions go up, it might be less likely to pay for carbon removal. The cost of saying “we are carbon neutral” is the cost per ton of carbon removal times the number of tons of carbon Microsoft emits on its own. If Microsoft is a software company — some people in an office typing on computers — its own emissions will be fairly small, and it can get the PR benefits of being carbon-neutral without a huge expense. If Microsoft is a vast network of power-hungry data centers used to build AI, it has a lot of emissions to offset, which is expensive. And the PR benefit of saying “we offset 60% of the carbon we emit” is, probably, not 60% of the benefit of saying “we are carbon neutral.”
2. As the PR benefit of saying “we are carbon neutral” goes down, so does the amount Microsoft should pay for carbon removal. The PR benefit, in the current US political environment, could be negative. If you are carbon neutral these days, you might keep that to yourself. I wrote a few weeks ago: “In 2026, if you did find a cheap clean renewable way to produce limitless fuel from algae, you’d keep quiet about it. ‘We made this fuel by burning coal in a new way that emits extra carbon, promise!’” 

Anyway Bloomberg’s Alastair Marsh reports:

Staff at Microsoft Corp. have told some developers of carbon removal credits that the company is pausing what is currently the world’s biggest program for financing the extraction of CO2 from the atmosphere.

Employees at the software giant have called a number of carbon project developers in recent days to say Microsoft is putting purchases on hold, according to two people familiar with the matter who asked not to be identified disclosing confidential information. In one instance, Microsoft employees said the decision was motivated by financial considerations, one of the people said.

Microsoft is by far the largest investor in removal credits, having set an ambitious goal to be carbon negative by 2030. The company is engaged in deals across a variety of technologies, with BloombergNEF estimating that its purchases in 2025 accounted for 96% of the entire market. …

While Microsoft has [previously] expanded its carbon removals program, the company’s greenhouse gas emissions have increased significantly on the back of its investment in data centers needed to power artificial intelligence. 

Right I mean if you keep spending more money on carbon removal and your net carbon emissions keep going up, that does seem like an expense you could cut?

I wrote yesterday about a difference between prediction markets and regular financial markets, which is that regular financial markets are all correlated with one another and prediction markets aren’t:

Financial assets tend to move together based on underlying economic factors; baseball games do not. There is no obvious reason that the result of the Mets game would tell you anything about the result of the Yankees game.

In a footnote, I added that “if you’re trading a factor model of sports bets please be in touch.” Someone emailed to point out that the result of the Mets game would tell you a lot about the result of the Yankees game if they played each other, which was not really my point.

Still, there are probably many domains where a bunch of prediction-market event contracts will all be correlated with one another, or maybe even with underlying economic factors. There are “mention markets” on what words Jerome Powell will say at a press conference, and I suppose you could build a macroeconomic model that predicts, like, Treasury bond prices and also Powell’s word choices. Or here’s a Wall Street Journal article about whatever this is:

On the day of President Trump’s State of the Union address, Foster McCoy sat himself in front of two glowing screens to “monitor the situation.” ...

McCoy, 28, was using all of these inputs to place a bet: that Trump’s youngest son, Barron, would show up in the audience.

“The only way people found out that Barron Trump was going was through Eric Trump’s Instagram Story,” the Minneapolis resident said. “If you followed him on Instagram, you found that he had posted a selfie with all of Trump’s children in one picture, all wearing suits, preparing for the State of the Union.” 

That one appearance, McCoy said, would set off a domino effect on other related markets, like whether Trump would shout out his kids during his speech, or use other specific words traders could bet on him using. His combined bets netted him about $10,000, he said. 

Obviously not a fundamental economic model or anything but, sure, I guess the point is that a lot of prediction-market bets are correlated to the factor “Barron Trump is on Eric Trump’s Instagram story.” It’s not that all events are correlated to the same fundamental factors, but every event that you can bet on is correlated to like 10 other events you can also bet on, so you need at least some intuitive model of how to trade related markets. Modern finance is so good.

Elsewhere, here’s a 2004 paper by Steven Levitt titled “Why Are Gambling Markets Organised So Differently from Financial Markets?” From the abstract:

In sports betting, bookmakers announce a price, after which adjustments are small and infrequent. Bookmakers do not play the traditional role of market makers matching buyers and sellers but, rather, take large positions with respect to the outcome of game. Using a unique data set, I demonstrate that this peculiar price-setting mechanism allows bookmakers to achieve substantially higher profits. Bookmakers are more skilled at predicting the outcomes of games than bettors and systematically exploit bettor biases by choosing prices that deviate from the market clearing price.

As we discussed yesterday, this is all a matter of degree: Financial market makers sometimes take positions, and sports bookmakers sometimes try to balance bets, but, yes, to the extent that prediction markets are essentially sports betting markets, their market makers will have to take directional positions.

JPMorgan Traders Blow Past Expectations With Record Haul. Citi Logs Best Returns in Five Years as Revamp Takes Hold. Wells Fargo Misses Loan, Fee Estimates Amid Lower Interest Rates. BlackRock Pulls In $130 Billion of Client Cash as Fees Boom. OpenAI investors question $852bn valuation as strategy shifts. Amazon to Buy Satellite Operator Globalstar for $90 a Share in Cash or Stock. United CEO Has Pitched Possible Tie-Up With Rival American. Meta Expected to Unseat Google as World’s Largest Digital-Ad Player. PwC plans overhaul of global consulting business. Young trader who made $250mn on Russian crude sets sights on Guyana. China Starts Bankruptcy Liquidation of Shadow Bank Zhongzhi. Fed Chair Nominee Warsh Discloses Assets Worth Over $190 Million. TCW Private Credit Fund Slashes Red Lobster Equity Value by 98%. AI chatbots misdiagnose in over 80% of early medical cases, study finds. Trump Jr.-Linked Cage-Fighters Make Pitch to Train US Military. Frozen yogurt fraud.

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Or you can subscribe to Money Stuff and other great Bloomberg newsletters here. Thanks!