Newsletter · · Ashutosh Agarwal

Private Credit Redemptions Slow as the Circularity Worry Grows - Private Credit & Alternatives - July 8, 2026

Private credit and alternatives newsletter for July 8, 2026. Blue Owl's semi-liquid funds saw redemption requests ease sequentially with no new non-accruals, a European CLO manager put software and AI at the center of the credit conversation, and the loudest bear reframed the risk from defaults to the circular financing of AI infrastructure.

Private Credit & Alternatives

July 8, 2026: Private Credit Redemptions Slow as the Circularity Worry Grows

Last week the story was a redemption snowball building quarter over quarter. This week the tape offered the first evidence it may be slowing, and more interestingly, the bears started arguing about a different risk entirely. The freshest hard number of the week came from Blue Owl's semi-liquid funds, where redemption requests eased sequentially. The freshest operator voice came from a European CLO manager who put the software-and-AI fault-line squarely at the center of the credit conversation. And the loudest bear on the tape spent his airtime not on defaults but on the circular way AI infrastructure is now being financed. Once again, the microphones that stayed empty are worth noting: no Apollo, Blackstone, Ares or Brookfield executive surfaced to defend anything, and there was still nothing from the insurance balance-sheet partners, the 401(k)/DC-access camp, NAV lending, or GP-led secondaries.

TL;DR

  • Redemptions may be peaking rather than compounding. Blue Owl's two semi-liquid vehicles both saw request rates ease quarter over quarter (the tech fund still ran hot at 38.1% but improved ~230 bps) with no new non-accruals in either fund. The stock had one of its best days ever on the read.
  • Software is now the consensus fault-line across the whole stack. A European CLO manager says even "conservative" tier-one managers are caught overweight software; Blue Owl is triaging: extending fresh financing only to software borrowers it judges can survive AI.
  • The bear case has shifted frame. David Rosenberg is less worried about a default wave than about the circular financing of AI, flagging the reported $35bn Blackstone/Apollo vehicle funding Google TPUs for Anthropic as late-1990s-telecom-style intertwining.

What's new

1. The redemption second derivative turned. The single most useful data point of the week came via CNBC's Leslie Picker on Squawk on the Street (Jul 2), reporting Blue Owl's disclosed redemption figures. Its tech-focused fund (OTIC) received requests totaling 38.1% and its broader private-credit fund (OCIC) 18.8%, both capped at the customary 5%. What moved the stock was the change: the tech fund's request rate improved roughly 230 bps quarter over quarter, with the broader fund also easing about 110 bps. Crucially, "there were no new issues in the quarter for either of these funds," credit quality held, no blowups. As Picker framed it, "since there is improvement… you're seeing kind of that work its way through the system and it's starting to really slow a little bit." One caveat on tape from the desk: we're only "two quarters" into this, likely too soon "where you'd expect to see actual erosion in terms of nonpayment and these companies going into default." Note the mechanical over-requesting she described: "it's kind of like a hot IPO… you put in more than you expect to get back."

2. Software is the fault-line, even for the "conservative" names. The sharpest operator commentary came from Accunia Credit Management's David Altenhofen, a European CLO manager, on The CLO Investor Podcast (Jul 7). His point on manager dispersion was blunt: some of the so-called "tier-one, conservative managers… have been overexposed to healthcare and software, and they have a little bit of issues right now." On the cycle: US CLO equity has had "a tough go" over two years (higher defaults, poor recoveries, loans repriced tighter, less income) "and then this year, to add insult to injury, you have the step up in AI capabilities, which could potentially put at risk some of the software business models you find in CLOs." His guiding philosophy is defensive: "It's not about choosing the winning managers. It's more about avoiding the losing managers." For scale, he cited a US CLO 2.0 market around $1.2 trillion (vs. ~€300bn in Europe), where only ~20-25% of deals are risk-retention compliant, and a 30-year CLO BB default rate of about 23 bps, "a de minimis default rate" that would rate between BBB and single-A on the corporate scale. Europe just logged its first recent CLO impairment.

3. The bear reframe: it's the circularity, not (yet) the defaults. On RiskReversal Pod (Jul 1), David Rosenberg tied private credit to the AI trade. His credit tell is classic: "credit always leads the equity market," and triple-C/double-B spreads "have widened out dramatically… that's usually a canary in the coal mine," and one company "after another gating their funds, capping redemptions" amid broad exuberance "raises my contrarian antenna." But the newer worry, teed up by host Dan Nathan, is structural: a reported ~$35bn Blackstone/Apollo vehicle set up to finance Google TPUs (Broadcom-designed) used by Anthropic, on top of Nvidia's vendor financing of its own customers. Rosenberg likened it to the late-1990s telecom bubble "when everybody was each other's customer," warning that hyperscalers ("historically… valued as being asset light and now they're capital intensive") are now burning cash and "compelled to come to the debt and equity markets to raise money… at the time when you're having some major lenders cap their redemptions." Treat the $35bn figure as an unverified assertion from the show, but note the direction.

4. A KKR operator pours cold water on the AI timeline. The useful counter to the software-doom came from KKR's Pete Stavros on Dry Powder (Jul 1). Across a portfolio "approaching 250 companies," running "dozens" of AI vendor experiments, his read is measured: AI is "helpful, but it is still a long way from transformational… it tends to be an incremental lever, not the driver of a deal outcome." On the wealth channel, he was equally deflating of the hype: the evergreen vehicle "has almost no impact on how we run the business" beyond funding certainty, it's a "pure co-invest" that takes a slice of every deal and mostly "displac[es] some of that non-client capital." Even so, "we're still calling hedge funds sometimes because we can't place the money." The structural tailwind is real: he cited that ~90% of companies with more than $100M in revenue are private, and the number of US and UK public companies is "down by half over the last 20, 25 years."

5. The street's rotation call. On the Squawk on the Street 9am hour (Jul 2), Jim Cramer and David Faber flagged a sell-side call to "sell Goldman and Morgan Stanley and buy KKR and Ares." Cramer's own lean: "I like KKR. I like Blackstone more," while pushing back on selling the banks into what he calls a "golden age of investment banking" and a coming M&A wave. On Blue Owl's 8.5% pop, he was unsentimental: "This stuff no longer means anything… you try to get your money out, you can't."

The debate

Is the redemption wave peaking, or is this the eye of the storm? The steel-man for peaking is now data, not narrative: Blue Owl's sequential improvement, no new non-accruals, capped structures behaving as designed, and retail visibly learning to over-request. The steel-man for the storm is timing: Rosenberg's spread-widening and the desk's own admission that two quarters is too little time to see defaults surface. Both sides actually agree on the where: software is the hole. The genuine disagreement is the when of AI disruption, and there the most credible operator on the tape, KKR's Stavros, says it is still incremental across 250 companies. If he is right, the software-loan fear is being priced early. That is the swing factor into 2H26.

What the silence still tells you. For the second week running, no mega-cap alt-manager executive appeared to defend gated flagships. There was no insurance balance-sheet-partner commentary (Athene, Corebridge, F&G and peers), nothing on 401(k)/DC access, and nothing on NAV lending or GP-led secondaries. In a week when the redemption data finally broke in the managers' favor, their continued absence from the microphone is its own small tell.

The names in play

  • Blue Owl (OWL): the quarter's data point: OTIC (tech, ~64% software) request rate 38.1% but improving ~230 bps; OCIC 18.8%; both capped at 5%; no new non-accruals. Management is triaging, extending fresh financing only to software borrowers it believes can "evolve alongside an AI-driven landscape." Stock had one of its best days ever on the read, though it remains down sharply over the redemption saga.
  • KKR and Ares (ARES): the "buy" side of the street's banks-to-alts rotation call. KKR's evergreen-as-pure-co-invest framing and Stavros's AI-is-incremental view cut against the doom narrative.
  • Goldman (GS) and Morgan Stanley (MS): the "sell" side of that same call, met with Cramer's pushback on the M&A tailwind.
  • Blackstone (BX) and Apollo (APO): named on tape only through the reported $35bn TPU-financing vehicle, the clearest sign the data-center/AI build is migrating onto private-credit balance sheets. Cramer prefers BX among the alts.
  • Accunia (private): a ~€2bn European CLO boutique running an "avoid the losers" book, a useful window into how disciplined credit shops are ranking managers on CCC and software exposure.

Read-throughs

  • Watch the second derivative, not the level. Blue Owl's ~230 bps sequential improvement is the first crack in the snowball thesis. If it repeats next quarter, the compounding-redemption narrative dies.
  • Software concentration is now the single tell across CLOs, BDCs and semi-liquid funds alike. That managers are selectively financing only "AI-survivable" software borrowers means triage (not denial) has begun.
  • Data-center and AI-infrastructure financing is the next thing to underwrite. The circularity Rosenberg describes (private-credit SPVs, vendor financing) is where the systemic worry is migrating.
  • AI-disruption timing is the swing variable. A KKR operator across 250 companies says it's still incremental; the CLO and doom camps say the software hole is real now. You cannot be right on private-credit software risk without a view on this.

What changed

For the first time this cycle, the redemption second derivative turned positive, sequential improvement at Blue Owl, with no new non-accruals. Just as notably, the locus of bear worry migrated from "private-credit default blowup" to "circular AI/data-center financing." And triage became visible: managers are now extending fresh capital only to the software borrowers they judge can survive AI, while credit shops rank the field to avoid the losers rather than pick the winners.