Autoresearch: per-name BDC exposure to AI-disrupted software/services borrowers

Generated by /autoresearch on 2026-06-17. Synthesized across 2 rounds (early-exit — per-name leg well-answered) from web search results; no Grokipedia anchor (fast-moving financial topic). Treat as raw material — review before promoting. Context: vault/projects/stock-market

Summary

This closes the per-name tradeable-endpoint gap that kept ai-capex-derate-to-private-credit-contagion a hypothesis: UBS's sector-level "25–35% exposed, not priced in" claim now resolves into a relative-exposure ranking of named, publicly-traded BDCs and managers. The split is tech-concentrated names with lagging marks (the underweight/short leg) vs. diversified blue-chips (the relative-winner/long leg). The load-bearing not-priced-in tell is concrete: Thoma Bravo's head of credit says software debt is trading in the secondary below where BDCs mark it — i.e. the markdowns observed in Q1 (software −7pts vs BSL-ex-software −1pt; 51-BDC aggregate unrealized loss 2.35% of NAV, steepest since Q2 2022) are an early installment, not the whole bill.

Findings

The mechanism, quantified (corroborates the UBS leg already cited)

UBS estimates 25–35% of private credit faces elevated AI-disruption risk, most acute in technology (~24% of BDC holdings) and business services (~30%); software is ~17–20% of BDC investments by deal count, second only to commercial services (PitchBook/UBS, Yahoo/UBS). In an "aggressive" disruption scenario UBS models private-credit defaults at 12–13%, vs ~8% for leveraged loans and ~4% for high-yield — private credit is the most exposed wrapper (Bloomberg, swissinfo). Asset-light software borrowers offer less collateral for recovery, so credit selection matters more than in prior cycles (CAIA).

The "not priced in" tell — marks lag the secondary

The sharpest piece of new evidence: Jeff Levin, head of credit at Thoma Bravo (a software-focused PE lender), said current conditions have opened secondary trading of software debt at levels below where BDCs mark those same loans in their portfolios (PitchBook, "talk of the town … Berlin"). That is the mechanical statement of "loans near par while equities/secondary diverge" — the gap that has to close through future markdowns. Q1 already showed the leading edge: software loans fell ~7 points vs ~1 point for the BSL market ex-software, and ~40% of unrealized markdowns were "Software" (PitchBook, Blackstone fund). A Reuters review of 51 BDCs put aggregate Q1 unrealized losses at 2.35% of NAV — the steepest quarterly hit since Q2 2022 (MarketScreener). Fitch's read: redemptions so far are sentiment-driven, not yet fundamental (PitchBook/CAIA) — i.e. the fundamental leg is still ahead, consistent with "not priced in."

The per-name ranking (the tradeable endpoint)

Most-exposed (relative — the underweight/short candidates):

• OBDC — Blue Owl Capital Corp. Software cut to 16% from 19% in Q1; management said it will "continue to be very cautious in software … look to continue to take that down"; dividend cut 16%, coverage barely matching NII, 11.3% yield (yield-as-distress-signal); Wells Fargo lowered PT to $12 from $13 (Equal Weight) (Seeking Alpha downgrade, TipRanks/Wells Fargo). The fact that management itself is actively cutting software is the cleanest admission of the exposure.
• OTF — Blue Owl Technology Finance (pure tech BDC). Trading ~67% of NAV, double-digit yield; hit a 12-month low on an analyst downgrade, Truist and Wells Fargo cut PTs (Defense World, Yahoo/Truist). Caveat to the short: credit metrics still strong — non-accruals ~10bps, 91.5% of assets at/above expectations — so the bear case here is concentration + sentiment, and one analyst frames the 67%-of-NAV as "buying the fear in the right place" (Seeking Alpha). The cleaner expression may be the manager, OWL, where Wells Fargo also cut the PT.
• BX — Blackstone (via BCRED, non-traded). Blackstone Private Credit Fund non-accruals rose to 2.4% at cost from 0.6% in Q1, driven by software and dental markdowns (PitchBook). BCRED isn't directly tradeable, but the read-through is to BX fee/FRE and to redemption pressure on non-traded vehicles (avg Q1 redemption requests 12.1%).
• GBDC — Golub Capital BDC. NAV dropped; software markdowns; loan terms turning more lender-friendly (a late-cycle tell) (PitchBook).
• FSK — FS KKR. ~$2.8B software exposure (Founderpath tracker).

Least-exposed (relative — the diversified/long candidates):

• ARCC — Ares Capital. Largest absolute software exposure (~$6.6B across 135+ companies) but the most diversified book — healthcare, financial services, consumer alongside software — with non-accruals ~1.8% late 2025; framed as a defensive core holding whose scale + low cost of capital wins on spread compression (Founderpath, FinancialContent). Diversification is the hedge — the long leg of a pair.
• BXSL — Blackstone Secured Lending. ~20% software but an exceptionally low ~49% LTV — conservatively structured relative to the exposure (Founderpath).

The trade structure this implies: a relative-value/pair — long the diversified blue-chips (ARCC, BXSL) vs underweight/short the tech-concentrated names whose marks lag the secondary (OBDC/OWL, OTF, GBDC, FSK; BX read-through via BCRED). This is within-cluster relative value, distinct from the equity-side seat-based-saas-ai-disruption chain — same root forcing function (AI eroding software economics), expressed in credit instead of equity.

Contradictions and open questions

• Sentiment vs. fundamental. Fitch says redemptions are sentiment-driven, not yet credit deterioration; UBS says the fundamental disruption is real but gradual. The short leg works only if the fundamental markdowns arrive — the Thoma-Bravo secondary-vs-mark gap is the best forward evidence, but timing is open.
• OTF is ambiguous. Strong credit metrics (~10bps non-accruals) at 67% of NAV cut both ways — concentrated-AI-risk short, or oversold "buy the fear." The manager (OWL) is the cleaner short than the vehicle.
• Most-exposed vehicles are partly non-traded (BCRED, some Blue Owl funds at 40–55% software) — the tradeable expression is the listed manager (OWL, BX, KKR) or the listed BDC (OBDC, OTF, GBDC, FSK), not the most-exposed fund itself.
• Does this graduate to a mechanism now? The per-name leg is cited and tradeable — ready for explore-chain to validate into wiki/mechanisms/ (the last gate in yesterday's log).