Private Credit Marks Fell $2.3 Billion in H1—Software Is Where the Losses Are Concentrating

Reuters' Sept. 4, 2026 review of regulatory filings from 44 U.S. business development companies found that portfolio fair value fell further below reported cost during the first half. At June 30, fair value was $92.88 billion against $95.19 billion of cost. At Dec. 31, 2025, the comparable figures were $95.82 billion and $96.54 billion. The gap therefore widened from $720 million to $2.31 billion. Marks deteriorated by roughly $1.59 billion in six months, even though outright defaults remain a separate and narrower measure of stress.

The Reuters review said losses were concentrated among over-levered horizontal software and services businesses with real exposure to artificial-intelligence disruption. That combination matters. A software lender is underwriting recurring revenue and enterprise value, not hard collateral. When investors question whether AI will compress prices, reduce headcount demand or shorten product life cycles, the same loan balance is supported by a lower enterprise value. AI uncertainty is turning an equity valuation reset into a debt-marking problem.

The apparent contradiction—low defaults alongside falling marks—is therefore explainable. Lenders can amend a loan, capitalize interest or wait for a borrower to recover while valuation models incorporate a wider spread and weaker exit assumptions. The loss becomes visible in NAV first and in cash collections later. BCRED's own Q1 disclosure illustrates the dispersion: its average portfolio mark was 96.4, while the bottom 5% of private debt investments averaged 69.6.

Blackstone Private Credit Fund received about $4.3 billion of repurchase requests in the third quarter, following $4.5 billion in the second quarter. The fund can repurchase shares equal to 5% of NAV each quarter. It fulfilled roughly half of Q2 requests, leaving approximately $2.3 billion unfulfilled; a significant portion was resubmitted in Q3. Redemptions are now testing the fund's promise of periodic liquidity, even though Q3 net outflows were reported at about 3% of NAV and nearly $750 million of subscriptions partly offset withdrawals.

This is the non-obvious transmission mechanism. A redemption gate can prevent immediate asset sales, protecting remaining holders from fire-sale pricing. But a persistent backlog also makes the fund's NAV less informative as a liquidity price: investors cannot freely exchange at that value. If software marks continue to fall, the queue can grow even while the fund technically meets its quarterly limit.

Public BDCs provide the cleanest listed read-through because they report loan marks, non-accruals and NAV quarterly. Ares Capital reported a Q2 2026 quarter ended June 30 and remains a large, diversified lender; its latest reported book value per share was $19.35 and its trailing dividend yield was about 9.6%. Blue Owl Capital reported Q1 2026 NAV per share of $14.41, non-accruals of approximately 2.0% at cost and 1.0% at fair value. High cash yields do not eliminate the risk of shrinking book value when stressed software loans are marked down.

The comparison is not a claim that one manager has the same software exposure as another; Blue Owl Capital's cited presentation did not disclose a specific software portfolio percentage, and Ares Capital's public figures do not isolate the software book in the materials reviewed. The defensible conclusion is narrower: the sector-wide mark deterioration raises the value of portfolio composition, valuation discipline and non-accrual disclosure. Investors should not treat a stable dividend as proof that NAV risk has passed.

The Financial Stability Board estimated private credit at $1.5 trillion–$2 trillion at the end of 2024, roughly comparable with the institutional leveraged-loan and high-yield markets. Banks provide subscription lines, revolving facilities and portfolio financing; about half of private-credit borrowers also maintain bank financing, usually revolving credit lines. For U.S. BDCs, as much as 90% of bank lending can take the form of credit lines. Shared borrowers create a second channel from software stress into bank credit exposure, although the FSB said separately identifiable direct exposures were generally below 0.5% of bank assets.

The CLO link adds leverage and opacity rather than an automatic crisis. Banks may hold senior, often AAA-rated, tranches and provide warehouse financing, while private-credit CLOs can contain a higher share of CCC-rated loans and weaker recoveries than broadly syndicated CLOs. The FSB also described possible risk loops in which banks fund nonbank lenders that buy bank synthetic-risk-transfer exposure. Those channels make transparency and collateral haircuts more important if software marks continue to weaken.

In the short term, BDC shares and alternative-asset managers are likely to move on redemption figures and quarterly NAV marks before software borrowers formally default. Over one to three years, the decisive variable is whether lenders' equity cushions survive an AI-driven change in software economics. BCRED reported that its software loan portfolio remained approximately two times covered by enterprise value after the sector sell-off, but that is a manager-reported cushion, not a guarantee of recovery.

The H1 2026 evidence supports a measured bearish conclusion for private-credit valuations, not a claim of imminent systemic failure. Aggregate BDC marks moved $2.31 billion below cost, software borrowers account for the most visible weakness, and BCRED's $2.3 billion redemption backlog shows that private valuations are being tested by actual investor liquidity demand. Software weakness is the trigger; redemption gates are the amplifier.

The most exposed trade is not simply “buy or sell BDCs.” It is a contest between attractive cash distributions and uncertain NAV durability. The bullish case requires stable software cash flows, limited PIK growth and a shrinking redemption queue. The bearish case requires only that marks continue to drift lower faster than subscriptions replenish the funds. Until Q3 filings show which path is taking hold, the prudent signal is dispersion: favor transparent portfolios and treat headline yields as incomplete compensation for valuation risk.