Excess Spread is our weekly newsletter, covering trends, deals and more in structured credit and ABF.

This will soon become part of our new subscription package, focusing exclusively on news and analysis within asset-based finance. Email marketing@9fin.com for more information.

Excess Spread is off next week

Excess Spread has been in Miami this week for ABS East. One vice in particular was the talk of the Fontainebleu, and that’s double-pledging, as the industry reels from the fallout of Tricolor and First Brands.

The words of the week were “idiosyncratic” and “cockroaches”, which describe the two schools of thought about these situations.

JP Morgan boss Jamie Dimon, whose august institution led many of the Tricolor ABS deals and financed the stricken lender’s warehouse, said last week that where there’s one credit cockroach, there tend to be more.

It’s a pretty tired metaphor, stale enough to have appeared as an Excess Spread headline about a different receivables blow-up just a month ago — but it seemed to catch the imagination this time, helped by big provisions taken last week by Western Alliance and Zions Bancorp on CRE debt, which seemed to reveal more cockroaches in real time.

So the cockroach people are looking for more bad deals to be revealed, while the idiosyncratic people think that fraud is specific and not endemic, and everyone should get on with their lives.

But as the examples mount up, “idiosyncractic” becomes increasingly difficult to sustain — though Blackstone president Jon Gray and Goldman Sachs's David Solomon are still on the team.

There are also different schools of thought on what to do about it. Many market participants take the view that “you can’t underwrite fraud”; a sufficiently determined bad actor in a position of corporate power can always pull the wool over lenders’ eyes.

What this actually means is “standard investment processes do not underwrite to a zero-fraud standard”. There is always more verification and checking that can be done, and each additional layer makes it harder to do certain kinds of fraud.

A lot of the things that stop frauds are pretty dull — it’s answering questions like “have we fully spelled out the process and governance by which loan samples are selected for additional spot checking” — and so it’s easy to skate by them in busy markets when the competition for deals is hot.

Some point the finger at private credit / private asset-based finance for helping to erode these standards. The more of the market that’s placed in private format, the harder it becomes to assess what’s market standard. Bilateral or club deals have fewer eyes on them than public markets.

That said, high yield is public-ish (9fin has the docs, at least) and that’s done nothing at all to prevent the erosion of covenant protections. If anything, second-tier sponsors that see how easy the big guys have it want the same flexibility in their deals.

Whatever else happens, there’s likely to be a focus on due diligence to avoid what is really a basic problem in securitisation. The first double pledging fraud in securitisation was probably, like, six months after the first securitisation, because it’s an extremely obvious and lucrative way to steal money.

It’s kind of amazing that in 2025 we don’t have a comprehensive solution to prevent it.

One possibility that might help is loan-level data. The industry in Europe howled when this was mandated, but has managed, with some fuss, to adjust to it. The ESMA-mandated disclosures don’t add a huge amount, and are very poorly tailored to investor needs, but it’s a nice idea in theory.

In practice, when VWFS wants to do a triple-A funding deal, it doesn’t get a whole lot from its mandated loan tape disclosure to the European Datawarehouse (it just has to do it), while the portion of the market where it could really make a difference (NPL/RPL portfolio sales) simply can’t get the data up to standard for pre-crisis legacy assets, and prices adjust accordingly. Yes, those automated valuations from 2007 are disclosed; no, they aren’t relevant any more; put your best foot forward anyway.

The US also mandated loan-level data post-GFC, but through Reg AB. This was downstream of US securities regulation, and consequently extremely easy to avoid by issuing 144A transactions. Thus there have been no public RMBS offerings since 2013 — public in this case meaning SEC-registered.

Private credit is just credit now

What if I told you about a debt instrument, issued under 144A documentation with a CUSIP and an ISIN, held in ETFs, with an S&P credit rating, issued to some of the world’s largest asset managers, trading in blocks of up to $500m, and marked to market in real time? How would you describe this deal?

If you answered “private credit”, award yourself a coconut. Of course it’s private credit! Everything is private credit! Just because it’s a highly liquid, tradeable bond, that doesn’t stop it being private credit.

I refer of course to Meta’s $27bn funding package for its under-construction Louisiana data center complex, Beignet Investor LLC. Really I should say Blue Owl’s funding package, because Blue Owl owns 80% of the project issuer and the debt is being raised against Blue Owl’s end of the investment.

That’s because this is an artfully structured off-balance-sheet instrument. In line with other private-credit, investment-grade transactions, this is structured to avoid appearing as Meta debt on balance sheet, though Meta effectively guarantees performance through minimum rental commitments, repair guarantees, residual value guarantees and more. That’s how it got to A+ with S&P; it’s essentially Meta credit risk in project finance clothing.

Considering Meta itself has only $29bn in unsecured debt outstanding (this is a large number in absolute terms but it’s a $1.8trn market cap) and the new facility pays 6.58%, you can see why there’s a certain eagerness to get involved on the part of the asset management industry. As of Monday it was quoted at 104.59, as of Tuesday 110.

The structuring of the transaction is pretty interesting — in every possible way, it’s structured as Meta credit risk, right up to the point where Meta has to consolidate the debt. The company does disclose that it expects an additional $52bn of lease obligations to come on balance sheet after the most recent June 2025 reporting date, some of which is likely the tenant obligations for this campus, but those are leases, not real debt, and Meta has four year break terms (which, admittedly, come with a residual value guarantee which will make the debt whole).

But we should spend a bit more time mocking the characterisation of this deal as “private credit”.

It was marketed by arranger Morgan Stanley as a private credit package, and the structure is massively more complex than an unsecured corporate bond. Certain purists could point out that all 144A deals are technically private placements to qualified institutional buyers, though that way madness lies (the entire HY and RMBS market is private credit? Sure!)

What private credit really means here is “optimised for off balance sheet treatment”. Meta could easily fund this investment through its unsecured shelf or even, with a little strain, from cashflow (it’s throwing off c. $100bn a year). But it doesn’t want to, and, although this project debt is considerably more expensive than where Meta funds in bonds, it’s worth it to separate the massively cashflow-generative corporate group from the monster capex required for its AI ambitions. It’s not optimising for cost of debt, it’s optimising for corporate structure.

Surely, though, this marks the top in terms of private credit = everything in fixed income? The term has been debased out of all recognition. How about we just call it “credit”, and make private credit private again?

Rational exuberance?

Meta’s campus is a monster development, but there are multiple data center mega projects announced or under construction in the US. Here’s $9bn from Google in Oklahoma or $20bn from Amazon in Pennsylvania or Microsoft in Wisconsin or, the most grandiose plan of them all, the OpenAI/SoftBank Stargate. I suspect we’ll see this kind of debt markets technology again.

Is it all a bubble though? The AI-adjacent equity markets obviously are, but at least if it’s an equity bubble it’s fun to be in. Debt bubbles don’t even have an upside.

Data center ABS, CMBS, or project financings like the Meta deal are an increasingly large part of the US securitized products market (data center ABS is now some 10% of the market).

Facilities for AI model training are the most eye-catching, but there’s plenty of plain vanilla data center investment to be done as well, which shows less bubbly characteristics. “Cloud computing” is pretty much just how the world runs these days, and it still needs data center investment.

The technicals are also fairly balanced. There’s plenty of capital available in private credit / structured products / bank lending to lend to data center deals, but this is also getting deployed in huge $27bn chunks, so there’s ample supply to match the apparent demand, especially if, like the Meta transaction, deals are priced to sell.

But it’s still an area that’s controversial among fixed income investors, whether they’re active in public asset-backed markets or somewhere in “private credit”. The history of data center performance isn’t much help as a statistical guide to the future, and you generally can’t model these as meaningfully diversified pools. The master trust structures typically used for US data center ABS give the owners lots of flexibility, but underwriting these is a lot more complicated than looking at, say, a credit card master trust; you need to get up the curve on chip types, power sources, different applications and business models, and understand what kinds of data center facilities can be added or removed from the collateral pool.

If it is a bubble, what pops it?

A lot of the specifically AI-linked investment is coming from companies, like Meta, which are massively cash-generative. If their shareholders are comfortable with them sinking funds into supporting AI capex (whether on- or off-balance sheet), then it’s not popping any time soon. The quality of earnings from Instagram ads, Azure contracts or Oracle subscriptions is a very solid backstop, provided it continues!

Excess Spread is our weekly newsletter, covering trends, deals and more in structured credit and ABS — subscribe to this newsletter here.