Gundlach found two junk bond markets and only the AI one is widening

Jeffrey Gundlach sat down with Julia La Roche on September 15, the day before the Federal Reserve met, and put the odds of a rate cut at zero. The Fed hiked 25 basis points the next afternoon to a target range of 3.75 to 4 percent, its first increase since 2023. That was the easy call. The harder one came earlier in the same conversation, when he took the high yield market apart and found that it has stopped behaving like one market.

He split high yield in two and only one half moved

The fight over AI capital spending has been an equity fight for two years. Gundlach moved it into credit and brought the arithmetic with him.

you're starting to see the non-AI sector is still strong. Spreads have barely widened on the junk bonds and their prices are near their peak on the bank loans. But when you look at the AI component, they've visibly, noticeably widened. The junk bonds are out about 50 basis points from their tights on the AI and the bank loans are out more like 130 basis points from their tights.

Jeffrey Gundlach, September 15, 2026, source

Fifty basis points on the bonds is a wobble. A hundred and thirty on the loans is something else. Loans float, so every one of those borrowers got repriced again the following afternoon. The widening has not reached the single B tier yet. That is the line he is watching.

So you'll notice that nothing in this mix has any like AI. There's nothing here. And I was perfectly fine owning some AI by using other types of equity vehicles. But starting last week, I just want out.

Jeffrey Gundlach, September 15, 2026, source

The whole 30 percent equity sleeve of his published allocation now sits in one equal weighted index of roughly 440 large US companies ranked by revenue. That is not a hedge against the AI trade. It is an attempt to hold nothing that touches it.

The ratings are the part he refuses to accept

His real complaint is not the spread level. It is that the letters on these bonds no longer tell him anything.

It's like when SpaceX borrowed a bunch of money, the bonds widened out to levels about three notches lower in credit quality. Interestingly, they got rated triple B minus, the lowest rating of investment grade. And I have a feeling that that rating was encouraged by some persuasion of the rating agencies because the bond market isn't buying it at all.

Jeffrey Gundlach, September 15, 2026, source

He has the detail wrong and the direction right. All three agencies graded SpaceX on June 18 and none said BBB minus. S&P said BBB, Moody's said Baa1 and Fitch said BBB plus. What followed was worse than his version. By early July the $25 billion deal was changing hands at an average 1.62 points over Treasuries against 1.55 for the BB index and 0.92 for BBB. The market did not mark it three notches down. It marked it through junk.

Then he told a second story, about an investor whose affiliated insurers reported 3 percent of invested assets in related party deals and later restated to near 42 percent, with half their ratings from one agency of 25 people that graded 3,200 deals in a year. Close enough to check. Prosecutors are examining Mark Walter's Delaware Life and Clear Spring Life, which restated related party private credit from $1.4 billion, or 3 percent of invested assets, to at least $17 billion, or 39 percent. The agency in the middle is Egan-Jones, which rated more than 3,000 private credit investments in 2024 with about 20 analysts, is under SEC investigation over whether commercial pressure shaped its ratings, and was dropped from Bermuda's list of recognized raters in December 2025.

So this is a very dangerous situation where the private credit is sort of the fuse and the insurance companies are the bomb.

Jeffrey Gundlach, September 15, 2026, source

Franklin Templeton is selling the trade he is running from

The same day, Jenny Johnson, who runs $1.7 trillion at Franklin Templeton and has spent six years buying her way into alternatives, gave the other side of it without being asked.

The regulators are not going to suddenly change the capital requirements on banks. Banks aren't going to be able to lend like they used to lend. So, the private credit's here to stay.

Jenny Johnson, September 15, 2026, source

Her number for what that pays is at least 100 to 150 basis points over public fixed income at investment grade, more as you go down the stack. She is also the one in this pair who keeps warning that the product is dangerous in the wrong hands.

So one is I'm a big believer that private markets should be delivered through financial advisors. It's a little bit of that running with scissors kind of thing.

Jenny Johnson, September 15, 2026, source

These two are not looking at different worlds. They agree the premium exists and roughly what it pays. They disagree about what the buyer is being paid for. Johnson prices it as compensation for locking your money up. Gundlach thinks a slice of it is compensation for a rating nobody read the file behind, which is a different risk with a much worse tail.

Where they converge is more interesting than where they split. Johnson's own stated worry is not credit at all. It is what the index turned into while everyone argued about active versus passive.

I'm not convinced the market has fully figured out how we're dealing with what used to be capital light businesses that are now becoming capital intensive businesses.

Jenny Johnson, September 15, 2026, source

That is Gundlach's equal weighted revenue index, reached from the opposite chair. One is running from the credit, the other from the concentration. Neither wants to hold what everybody owns.

The Fed made the floating rate borrowers worse off the next day

Gundlach talked himself into the hike off a Jackson Hole anecdote about hiking, which is a thin reason to move a forecast, and he said so. Warsh went further at the press conference than the statement did, saying the plain fact is that inflation is too high and has been for too long. The median official now sees one more hike this year.

His harder claim is about what prints next. August CPI came in at 3.4 percent, with gasoline up 27.4 percent over the year and responsible for a third of the monthly move.

We think that the next print that comes out on the CPI will start with a four based upon the oil movement and stay above four unless something changes with the commodity complex stays above four all the way through March.

Jeffrey Gundlach, September 15, 2026, source

Set that beside the loan data and the sequence gets unpleasant. Triple C loan borrowers float, and most of them are waiting for a cutting cycle to refinance into. A four handle on headline CPI pushes that cycle further out than any of them underwrote.

Washington is reading the safety fight as a financing story too

The week's loud argument was whether the frontier labs should slow down, which I covered in the pacing essay and the sellers' response. What is new is how many people now read it through the cap table. Emil Michael, the Pentagon's research and engineering chief, said it out loud on CNBC on September 16.

Elon softened his comments and said it was 4D chess because of the IPOs coming up for some of these companies. And maybe there's a relationship between some of their comments on slow pacing the frontier with their financial interests.

Emil Michael, September 16, 2026, source

Asked about nationalizing any of it, he answered in valuations. He named 1.25 trillion, 1.5 trillion and 2 trillion, and called them potentially two of the top five biggest companies in the history of the world. Gundlach, who has no stake in the safety argument, noticed the same calendar and was blunter, pointing at a dormant account that drew 150 million views in hours and the timing seems awfully strange relative to the midterms, relative to an IPO that was being organized for maybe October.

Which raises the thing the credit side has not priced. Jeremy Allaire told Nicholas Thompson on September 16 that the firms coming to market are a different animal from the comparables.

we're entering an age where essentially capital is deployed through agents and accrue more capital without humans. You create potentially dramatically more capital concentration, right, than we've had.

Jeremy Allaire, September 16, 2026, source

He put Anthropic at a possible $100 billion revenue run rate by year end, up from 65. If that is even half right, the borrowers Gundlach is avoiding are growing into their debt faster than anything in his history.

So which of them is wrong about private credit?

Johnson is, on the one point she conceded herself. She said the $40 trillion of federal debt is not a problem right now. Gundlach's case is that the private ratings already stopped working, and what exposes it is a rate path that stays above what borrowers assumed. The Fed just raised into 3.4 percent inflation and guided to another. That is not a future problem. It lands on the floating rate part of the market this quarter.

The number worth keeping on a card is 130 basis points. If AI bank loans give most of that back and the rest of the market holds, Gundlach sold a dip and Johnson collects her premium. If the gap spreads into single B, the argument about whether the models are dangerous becomes a far smaller story than the argument about who wrote the ratings.