The Calm Is in the Wrong Place

Two pictures of the same market, side by side

Start with the two facts, plainly, because the contrast is the whole point. On one screen: the public bond market is calm to the point of serene. Investment-grade credit spreads — the extra yield lenders demand over Treasuries to hold corporate risk — sit near 75 basis points over Treasuries, close to their tightest of the cycle. And SpaceX, a company that lost roughly $4bn last quarter and burns cash, just sold its first bonds with investment-grade ratings from all three agencies and an order book of $89bn — demand so heavy it let the company tighten the price.

On the other screen: the private side is bolting the doors. In the space of one month, three large managers — Blackstone’s flagship credit fund, Partners Group, and now Apollo — have capped how much investors can pull out of their retail credit vehicles, after withdrawal requests jumped (Apollo’s reached almost 17% in a single quarter). One market is being offered more money than it knows what to do with; the other is being asked for its money back. Both are ‘credit.’ The instinct is to ask which one is right. The better question is why they can both be true at once.

The risk didn’t vanish. It moved to where the money is calm.

The reason is that the way the AI build-out is being financed pushes risk upward — toward the calm, rated, public top of the capital structure — and away from where anyone is looking. It happens three ways, and SpaceX is the live illustration of the first.

One: going public manufactures a credit rating. A loss-making ‘startup,’ however large, cannot easily sell investment-grade bonds. A loss-making public company can. Once there is a listed share price and a crowd lining up to own it, the rating agencies lean on that — SpaceX’s ratings explicitly cite its access to capital markets and the promise of its AI segment, alongside real Starlink cash flow, even as one agency expects it to burn cash through 2029. The listing did not just raise equity; it created the capacity to borrow. Note the tell, though: the 10-year bond still priced around 140 basis points over Treasuries — roughly 60bp wider than the investment-grade market average, and wide even of a typical BBB-rated borrower. The market said yes — but charged a premium for the novelty and the losses. Heavy demand and real wariness, in the same deal.

Two: a real balance sheet gets rented. A data-centre borrower without the earnings to carry its debt can still borrow cheaply if an established investment-grade company stands behind it — someone else’s rating carries the loan. The live example: Meta’s ‘Hyperion’ campus was placed in a separate venture that raised roughly $27bn of top-rated debt — for a brand-new entity — because Meta signed the lease and added a residual-value guarantee: a capped promise to cover the lenders if the lease ends. They were lending against Meta, not the project. The precision matters: that is real recourse to the marquee name, but short of a full guarantee of the debt — it caps the loss, it does not erase it.

Three: the leverage hides off the balance sheet. Much of the build-out’s cost is rented — compute, power, data centres committed for years — so the obligation flows through as leases and purchase commitments rather than as debt. The borrower can look lean while the real claim on its future cash is enormous. So the public calm is partly manufactured. Spreads are tight in part because the riskiest pieces have been relabelled, anchored, or moved off the page where a credit spread can see them. The calm is real — it is just looking at the wrong place.

The crack appears where the least patient money sits

If risk floats up to the calm layer, where does the stress show? At the bottom — in the retail ‘semi-liquid’ funds that let ordinary investors put money into private credit and ask for it back each quarter. That is the most impatient money in the system, and it is the first to head for the exit. Three gates in a month is that exit getting crowded.

Here is the part worth holding onto, because the headlines miss it: a gate is a feature, not a failure. These funds are allowed to cap withdrawals precisely so a manager is never forced to dump assets into a thin market to pay leavers. The cap is the shock absorber doing its job. So far this is a liquidity event — people wanting out faster than the fund can sell — not a solvency event, where the underlying loans actually go bad. The distinction matters enormously, and it is the discipline this piece wants to keep: separate ‘hard to get your money out today’ from ‘the money isn’t there.’ They are not the same thing, and the gates so far point to the first.

And you do not have to take ‘liquidity, not solvency’ on faith. One slice of private credit is public and priced daily — the listed business development companies (BDCs), which lend to the same mid-sized private borrowers and publish their loan books each quarter. They are the window into the borrowers beneath the gates. Their just-filed first-quarter numbers say three things:

Read together: the borrowers are straining — deferring cash interest, a handful tipping into non-payment — but they are not yet failing, and the funds that hold them are sound. That is exactly a liquidity event with a solvency question building underneath it. The PIK creep is the early-warning line; the non-accrual rate is the one that decides whether ‘not yet’ becomes ‘now’.

And ‘now’ is not hypothetical — we have just seen its shape. When a private loan goes bad it tends to go one of two ways. The orderly way: in April the owner of Medallia, a software firm, handed it to its lenders in a $5.1bn debt-for-equity swap — the owner’s stake wiped out, the lenders taking the keys and putting in fresh money to keep it going. A real loss, but worked out calmly. The messy way: last autumn’s sudden collapses of First Brands (car parts) and Tricolor (subprime car loans), where hidden borrowing meant the loss landed all at once. The gated retail funds are nowhere near either — but the clean-up is already under way in software and car loans, which is why the trend in PIK and bad loans is the thing to watch, not the gates themselves.

Does the stress travel upstairs? (the question that actually matters)

The real question — the one worth watching — is whether the stress travels: from the impatient retail money that is being gated, up to the patient, long-dated, institutional money that has no need to sell. The reassuring answer is that gates protect that patient pool: they stop forced selling, full stop. The honest answer is that they protect it from selling — not from repricing. And there are three wires along which a repricing can travel.

The marks. When a gated fund finally has to value or sell an asset, it re-prices the same loans sitting in other funds and on other balance sheets — so a patient holder who never asked to leave can still watch the value of what it owns move. The bank channel. These funds borrow from banks against their assets; if a bank marks that collateral down, it can call for more cash — a demand that hits every investor in the fund, patient or not. The sponsor. A manager seen gating its retail fund finds its next big institutional fund harder to raise, which can force it to shrink at the worst moment.

So here is the two-sided close. One reading: the gates are working exactly as designed, the system is absorbing a rotation out of an over-sold retail product, and the patient pool rides it out untouched — a healthy stress-test passed. The other: a bank-channel margin call or a forced mark turns a paper repricing into real selling, and the calm public layer discovers it was pricing risk it could not see. You do not have to guess which — you watch for it: a fourth manager gating, the first real markdown in the bank channel, and whether the recent wobble in stocks deepens to meet the weakness in private credit rather than ignore it. The patient pool is protected from having to sell. It is not protected from being repriced — and the line between those two is the whole game.

What to do with this — and what we are not doing

What to look for. Whether a fourth manager moves to cap withdrawals; the first real write-down in the bank channel (a lender marking fund collateral lower, not just a fund slowing exits); whether the recent softness in equities deepens to meet private-credit weakness rather than shrug it off; and — in the public BDC window — whether this quarter’s PIK creep starts turning into a rising non-accrual rate, the point at which ‘stress deferred’ becomes ‘loans gone bad.’ Those are the tells that a contained, by-design rotation is becoming something that travels.

What to be wary of. Reading tight public spreads as an all-clear — this piece is the argument for why that calm can be blind; reading a redemption gate as a collapse — so far it is the shock absorber working; and treating “private credit” as one thing, when the retail semi-liquid pocket and the patient institutional pocket are behaving very differently under the same headline.

What we are not doing. This is general market commentary, not advice or a recommendation on any security, fund or manager. The names here are illustrations of a mechanism, not calls. Whether any of this touches your own portfolio — your credit exposure, your liquidity, your tolerance for being repriced — is a suitability conversation, one that needs a person, not a page. We are always glad to have it.