The Gap That Tells You Who Is Right in Private Credit's First Run

On 23 June, Apollo told the investors in its flagship retail credit fund that they could have just under a third of their money back. They had asked for far more. Shareholders in the Apollo Debt Solutions BDC requested to withdraw 16.8% of the fund, about $2.4bn, in the second quarter, the largest redemptionredemptionAn investor asking for their money back out of a fund. request since the fund launched in January 2022. The fund holds roughly $14bn in net assets (about $27bn once you count the money it has itself borrowed), and redemptions are paid out of those net assets, which is why 16.8% comes to $2.4bn. Apollo held the line at its 5%-a-quarter cap and will honor less than a third of the demand on a proratedproratedSplit proportionally; when a fund can only honor part of the withdrawals requested, each investor gets the same fraction of what they asked for. basis.

On the same day, Morgan Stanley capped a $7bn fund of its own after investors asked to pull 11.6% of its shares. And three weeks earlier, on 4 June, Blackstone had capped its $79bn BCRED fund for the first time ever, after requests hit 10%. The biggest names in the business hit the brakes within about three weeks of one another.

This is the first real stress test of a structure that barely existed a downturn ago: the semi-liquid business development company, or BDC. Think of it as a fund that lends directly to mid-sized companies (loans that are illiquid and almost never trade) while promising ordinary investors they can cash out every quarter. It is a bank's core balance-sheet trick, borrowing short to lend long, dressed in a mutual-fund wrapper and sold to retail. The broader private-credit market it belongs to has grown to somewhere around $1.5-2.0tn, most of it built since 2020.

So the funds are gated. The interesting question is not that people are leaving. It is what their leaving tells us.

When investors line up to redeem far faster than a fund admits its loans have fallen in value, one of two very different things is happening.

• A confidence run. The loans are basically fine, but a semi-liquid fund can be emptied faster than it can sell illiquid assets, so the rational move is to be first in the queue regardless of fundamentals. Scary, self-curing, no real losses at the end.
• Front-running the truth. The published NAV (the fund's own stated price for its loans) is stale fiction. The loans are worth less than the fund says, and the people leaving are simply the ones who can see it first. Not a panic. A correct read.

From the outside these look identical. Both produce gates and ugly headlines. But they end in opposite places: one round-trips to roughly zero loss, the other ends in restated NAVs and real holes. Everything else in the news bundle, the AI-driven software markdowns, the bank exposure, hangs on which one this is.

The way to tell them apart is to measure the gap between how fast money wants out and how far the funds say their loans have fallen, and then ask which side of that gap is telling the truth.

Here is the wedge, using the freshest quarter on record. One bar is how much investors asked to withdraw in Q2; the other is how far the funds say their NAV has fallen from its peak. A word of caution before reading it: these are different units. The withdrawal figure is a flow (how much of the fund was requested in three months); the NAV figure is a price change. They are not directly comparable, but the size of the gap between them is exactly what the two stories disagree about.

Money wants out two to three times faster than the loans admit to falling. In Apollo's case, more than three times. And the run is getting bigger, not smaller: Apollo's request jumped from 11.2% of shares in Q1 to 16.8% in Q2; BCRED's went from 7.9% (which it fully honored, with senior executives reportedly helping fund the payouts) to 10% (which it gated).

There is one more tell hiding inside Apollo's number, and it cuts against the simplest panic story. The surge did not come from the US retail investors the BREIT comparison would predict. Onshore US requests actually moderated to 4.3% of shares in Q2. The pressure, 12.5%, came almost entirely from Apollo's offshore channel. That channel mixes private-wealth and institutional money rather than being purely retail, so it is suggestive rather than conclusive, but it is the opposite of what a headline-driven retail stampede looks like.

The natural reference point is BREIT, 2022-2023, Blackstone's non-traded real estate trust, the same perpetual, semi-liquid, retail-facing structure. When interest rates spiked, redemption requests blew through its 5%-a-quarter gate for more than a year while its appraisal-based NAV stayed marked far above where publicly traded REITs were valued. Commentators predicted a doom loop of forced selling. It never came. Requests faded, the gate came off in early 2024, and the feared losses largely round-tripped.

So the base rate says: bet on the round-trip. If this is BREIT again, the redeemers are wrong, the NAV is roughly right, and the gates will quietly come off in a few quarters. To overturn that prior, you need evidence that the loans themselves, not just investor nerves, are deteriorating. So look at the loans.

BREIT's defense was that nothing was actually breaking; the assets kept paying and the panic was about interest rates, not about the buildings. Private credit cannot make that claim today. The thing the round-trip story needs, quiet fundamentals, is missing.

Start with realized defaults. Fitch Ratings put the US private-credit default ratedefault rateThe share of loans where the borrower fails to meet the terms, such as missing payments or restructuring. at a record 6.0% for the 12 months to April 2026, up from 5.7% in March. That is roughly 2.4 times the 2-2.5% rate that counted as normal before 2025. Of 99 default events in the period, 81 were first-time defaulters, the most since Fitch began tracking in August 2024. This is not a stable book of loans being mistrusted by jumpy investors; it is a book that is genuinely cracking.

Now the mechanism that keeps the cracks off the books. Fitch found that more than half of those defaults were driven by interest-payment deferrals and the switch to payment-in-kind, or PIK, where a borrower stops paying interest in cash and instead piles it onto the loan balance. Picture a tenant who stops paying rent and instead signs IOUs, while the landlord keeps recording the rent as collected. A loan that has gone PIK can sit at close to its original value on a fund's books even as the cash stops arriving. That is precisely the gap between a stale NAV and reality, and here it is the dominant form of distress, not a footnote.

Finally, the same credit risk priced two ways. Between October 2025 and February 2026, public software stocks fell almost 30% on fears that AI would hollow out the subscription-software business model. Listed BDCs, which hold the loans to many of those same companies, fell about 10% on average, and the BDCs with the heaviest software exposure fell about 5 points more than that. Yet inside the semi-liquid funds, reported NAV is down only about 5%. When the public market and the private appraisal disagree this loudly about the same borrowers, the appraisal is usually the slower of the two to tell the truth.

Why is software the fault line? Because direct lenders leaned into it hard. According to the Bank for International Settlements, lending to software-as-a-service firms grew from under $8bn in 2015 to more than $500bn by the end of 2025, about 19% of all direct loans. BDCs specifically put more than 15% of their loans into the sector, and a third of all private-credit funds now have some SaaS exposure.

That concentration is the reason the public software sell-off matters so much for these private books. The funds did not lend a little to a fashionable corner of the economy; they lent enough that a re-rating of software credit is a re-rating of the funds. And the trigger for the public re-rating, the fear that AI will erode subscription-software revenues, is exactly the kind of slow, structural threat that an appraisal-based NAV is poorly equipped to register in real time.

If this were only about a few wealthy investors waiting an extra quarter for their money, it would be a footnote. It reaches further because private credit is wired into the banking system. Banks lend against the capital that investors have committed to these funds (so-called capital-call or subscription lines), and they finance the funds' own borrowing. The Financial Stability Board flagged exactly these linkages in its May 2026 review of private-credit vulnerabilities.

The danger is procyclicalityprocyclicalityThe tendency of a financial mechanism to amplify a downturn, for example forced selling that pushes prices down further just as everyone wants out., the tendency of a mechanism to amplify a downturn. If gated funds are eventually forced to sell loans to meet redemptions, they sell into a falling market, which validates the lower public marks, which deepens the discount, which encourages the next wave of redemptions. The gates exist precisely to interrupt that loop. They buy time. What they cannot do is turn an overvalued loan back into a sound one; if the NAV is the number that is wrong, a gate only postpones the reckoning.

Put the evidence side by side. The comfortable base rate, drawn from BREIT, says runs like this round-trip and the redeemers end up looking foolish. But BREIT round-tripped because its assets were sound and only its investors were scared. Here the assets are the problem: defaults are at a record and still climbing, more than half of them are being hidden inside PIK and payment deferrals that keep loans marked near par, the public market has marked the same credit risk down six times as far as the funds have, and the investors heading for the door are disproportionately the offshore, less retail channel rather than the panicky retail base the base rate assumes.

None of this proves the redeemers are right about the eventual losses; we will only know that as loans mature or restructure. But it tilts the balance hard. The redeemers are closer to right than the soothing base rate suggests. This looks less like a confidence run on healthy loans and more like a slow solvency event still wearing liquidity clothing, and the reported markdown, not the redemption wave, is the number that is lying.