How Loans Get Sliced Into Tranches

Three hundred and twenty-five billion dollars. That's what the safest slice lost in the crash — the triple-A, the top of the pile, the bit that was never supposed to lose a cent.

And people bought that because it was rated safe.

Right. Now — at that exact same moment, Wall Street was running a second machine. Same legal box, same waterfall, same rating agencies signing off.

And that one lost what?

Nothing. Zero. Not a dollar of principal. Not then, not since.

Same parts, opposite ending? What's actually different?

One number. You'll get it, but you need the machine first. So — somewhere right now there's a company you've never heard of that makes industrial valves. It borrowed money. That loan got sold to a shell company in the Cayman Islands, which owns about three hundred other loans just like it... and the interest that valve company pays every quarter ends up in a Japanese bank's bond portfolio.

Through how many hands?

Fewer than you'd think. That's the trick. The box is a legal thing — a special purpose vehicle, built to do nothing but hold loans.

Why a box? Why not just... own the loans?

Because the box lets you cut the money up. Cash flows in every month from borrowers, and instead of splitting it evenly you build a waterfall. Top layer gets paid first — triple-A. Then the middle, double-A down to double-B. At the very bottom sits the equity, which isn't rated at all and has no coupon.

No coupon? Then what do they get?

Whatever's left when everyone above them is paid. Could be twelve percent. Could be nothing. And losses run the other way — upward. Borrowers default, equity eats it first. In a typical deal about thirty-seven percent of the structure sits underneath that triple-A.

So the guy at the bottom is a human airbag for the pension fund at the top.

Paid extremely well to be an airbag. And everyone wants a different slice. Banks and insurers take the triple-A because regulators let them hold it cheaply. Hedge funds take the equity because they want the leverage.

And the alphabet soup — R-M-B-S, A-B-S, C-L-O. Same box, different label?

Different contents. Mortgages on houses, residential. Offices and malls, commercial. Asset-backed is everything consumer — car loans, credit cards. And then the esoteric corner, where people have securitized music royalties. Data centres.

Hold on. Data centres. We're coming back to that.

Oh, we are. But first the C-L-O. Collateralized loan obligation — a pool of loans to real companies. Mid-sized, already indebted, below investment grade. A hundred and fifty to three hundred and fifty borrowers in one deal. And here's the bit that matters: somebody actively manages it. A human can sell a loan that's going bad.

Which the mortgage pools couldn't.

A mortgage pool is frozen the day it's born. So, two thousand seven. Wall Street had run out of good mortgages. So they took the worst slices of subprime deals — the triple-B bits nobody wanted — pooled THOSE, and rated the top of that pool triple-A.

Sorry. You stack up the garbage, and the top of the garbage is gold?

That's the alchemy. And it rested on one assumption. Correlation. Michael Lewis reports that Moody's and S and P judged those pools of triple-B bonds to have a correlation of around thirty percent. Meaning if one went bad, the others mostly wouldn't.

But they're all subprime mortgages. Same bet, different envelopes.

Written in the same few years, by the same lenders, under the same collapsing standards, on the same national belief that house prices don't fall everywhere at once. The real correlation was close to one.

So the diversification was fictional.

And then they went further. C-D-O squared — a C-D-O made of other C-D-Os. Deals built on top of those. Bets layered on bets. That's where your three hundred and twenty-five billion comes from: by Oaktree's accounting, triple-A rated C-D-Os issued before two thousand eight. Now the number I promised. Standard and Poor's found the triple-A tranches of C-L-Os issued before the crisis lost nothing.

Still zero.

And according to a Meketa primer, no triple-A C-L-O tranche has taken a principal loss since the product was invented in the early nineties. Thirty-odd years. Multiple recessions.

Why, though? Genuinely. Why does one vaporize and the other shrug?

Because a valve maker in Ohio, a hospital group in Texas and a software firm in Germany don't all fail the same week. Real diversification, actively managed, no re-securitization. The structure was never the villain. The correlation assumption was.

So everyone learned the lesson and we all lived happily.

The C-L-O market is now over one point three trillion dollars globally, inside a structured credit market of roughly thirteen trillion. More than fifteen hundred live deals. And C-L-Os are the marginal buyer for about two-thirds of all broadly syndicated loans — so if that market sneezes, mid-sized companies stop getting credit.

Okay. The data centres.

Sit up for this. Data centre securitizations went from about four billion dollars outstanding in twenty twenty to around sixty-one billion by mid twenty twenty-six, per Barclays. Issuance hit twenty-six billion in twenty twenty-five alone — more than ten times the twenty twenty level. And the S E C has clarified that certain data centre deals aren't subject to some of the risk retention and disclosure rules other A B S face.

So the cash flow behind those bonds is companies renting server space to train A-I models.

One bet. One sector. One story about the future. Sound familiar?

Everybody's house goes up forever.

The Bank of England's Financial Policy Committee has flagged three things about private markets: widespread leverage, opacity of valuations and reliance on rating agencies, and correlation with other risky funding markets. Read that list again. It's the two thousand seven list.

That's uncomfortably tidy.

C N B C reported this year that private credit's two trillion dollar boom has watchdogs worried — the Bank of England running stress tests with the industry, deputy governor Sarah Breeden pointing at asset quality, valuation discipline, liquidity.

And what do I do with any of this? I don't own a tranche.

You probably do, through a pension or an insurer. But the useful thing isn't the ownership, it's the question. All of this is a bet that bad news won't arrive all at once. So when someone tells you a pool is safe because it's diversified, ask one thing. Diversified how? Different borrowers... or different names on the same story?

And if it's the same story?

Then there's no top of the waterfall. There's just the drop.