Recoveries on defaulted leveraged loans have fallen from a 25-year average of about 60 cents on the dollar to roughly 33 cents, Bob O'Leary said — a sign, he and Armen Panossian argue, that private credit's problems are deeper than the headlines about a few blown-up borrowers suggest.
Most coverage of direct lending's troubles has focused on isolated fraud cases and software-sector defaults. O'Leary and Panossian, the newly appointed co-CEOs of Brookfield's credit business, say the real story is structural: a 2021-2022 vintage of loans that's already producing three-quarters of this cycle's defaults, and a similar leverage pattern now showing up in the loans financing AI data centers.
"We do think there are a few problems and we want to bring up what those are and how we think we are addressing them and frankly going to profit from them over time."
O'Leary co-runs Oaktree's Global Opportunities strategy and Panossian its performing-credit business; both were named co-CEOs of credit for Brookfield, which completed its full acquisition of Oaktree this year, putting them in charge of underwriting across the direct lending, distressed and structured-credit portfolios now facing the vintage they describe.
I listened to the full episode so you can skip it. 34 minutes of audio, 22 minutes of reading.
Here are the 9 takeaways that matter.
👤 Guests: Bob O'Leary, co-CEO of Oaktree's credit business and portfolio manager for Global Opportunities, and Armen Panossian, co-CEO of Oaktree's credit business and head of Performing Credit
🎙️ Host: Harry Whitelaw, vice president on Oaktree's Marketing & Communications team
📰 Published: 9 September 2026, on Oaktree's own site
🔴 YouTube | 🟣 Apple Podcasts | ⏱️ 34 min | ✅ Time saved: 21 min
Key Takeaways
Defaulted-loan recoveries have fallen from a 25-year average of about 60 cents on the dollar to roughly 33 cents
O'Leary calls it a dramatic change from historic underwriting
Three-quarters of this cycle's defaults trace to loans issued in 2021 and 2022, and the real wall of maturities hasn't hit yet
Those loans mature in 2028 and 2029; any borrower that could refinance already has
Software went from private credit's favorite sector to its biggest liability once AI made customer end-markets correlated
Exposure ranges from 3-5% of high-yield to as much as 70% in some BDCs
The AI buildout needs $5.5 trillion to $7 trillion over five to eight years, and roughly $3 trillion of that has to come from credit markets
Spreads on data-center loans have started widening in recent weeks
Semi-liquid BDCs have hit their 5% quarterly redemption cap, and the managers who kept the most liquidity are the ones able to buy now
LPs are bundling their private-credit stakes and selling them on the secondary market rather than wait out promised 7-year fund lives
The distressed-debt universe — yields above 15%, prices under 90 cents — stands at about $250 billion, mostly a function of higher rates rather than a credit cycle already turning
Some BDC managers are marking a position at par one quarter and to zero the next, and there's no rulebook that says they can't
1. Four Themes, One Backdrop
Whitelaw opened by asking O'Leary to run through the themes Oaktree has been watching. He named four, building on the "Sea Change" thesis Howard Marks wrote in December 2022.
Rising Treasury yields are the first and most persistent theme. Two-year Treasuries are up about 80 basis points since the start of the year, which O'Leary attributed to relentless issuance, a couple of hot CPI prints earlier in the year, and a Fed that appears to be shrinking its balance sheet
The consumer looks fine in aggregate and isn't underneath it. "You're seeing good in the aggregate consumer numbers, but a lot of that's being driven by the highest end," O'Leary said, calling the divergence between high- and low-end spending an increasingly important backdrop to credit
Dispersion has only widened since Oaktree last discussed it. Triple-C loans trade at almost 2,000 basis points of spread versus about 500 for the overall loan index
Direct lending's problems are real, and O'Leary named four of them: software, fraud, rising defaults with worse recoveries when they happen, and — specific to BDCs — a mismatch between illiquid assets and shorter-duration liabilities. "I don't think either Armen or I are here to tell you that there's not a problem," he said, adding that fraud cases from late 2025 have quieted somewhat: "And I think while those headlines have calmed down a little bit, there is still a bit of a drip feed of that out there."
2. A BDC's Liquidity Squeeze
Whitelaw asked O'Leary to zoom in on the business development companies, or BDCs, that hold much of this direct-lending exposure.
Semi-liquid BDCs promise investors up to 5% quarterly liquidity, and most of the sector has hit that ceiling in recent quarters. Their assets are illiquid private loans; their liabilities are shorter-duration revolvers and bonds
Managers who kept more liquidity going in are the ones able to act now. "The most important takeaway from this is that when you've managed towards greater liquidity leading up to this event, then you're able to become more opportunistic through this period of time," O'Leary said. A manager with little liquidity also has less room to add bank leverage to fund redemptions
The levers to raise cash, in order, are cash on hand, liquid instruments like broadly syndicated loans or high-yield bonds, undrawn bank revolvers, and finally selling private positions at a discount. BDCs have already sold portfolios this way earlier in the year
Forced sellers can't sell what they'd like to. A BDC overweight software would ideally sell down that exposure as redemptions concentrate its remaining book further into the sector. "So ideally, you sell software, but the market today is not accepting of software transactions." There's little new issuance and little par trading in software loans
The effect compounds: a BDC without the liquidity to buy into today's wider spreads sees the average spread and equity performance of its remaining portfolio diminish for investors who stayed
Redemption queues at some BDCs imply years of withdrawals, not a quarter or two, which O'Leary said keeps sentiment risk hanging over specific managers for an extended period
3. The Vintage That Can't Refi
Whitelaw pointed to a chart of 2021-2022 vintage loans, which O'Leary called one of the worst vintages ever.
Private equity firms mistook a COVID-era profit spike for a structural change and levered accordingly. "They saw profit margins really expand during the COVID era restrictions. And I think they took that as a structural change in the industry when in fact it was really just borrowing forward demand," O'Leary said of building-products companies specifically, several of which have gone from record profitability to zero or negative EBITDA
Chemicals, packaging, building products and media are the sectors under the most pressure, for reasons ranging from reshoring-driven oversupply in chemicals to a capex arms race in media
Liability management exercises mostly just delay the default, and delay hurts recoveries. "Very few of them have gotten to the point where they're able to successfully refinance the debt and live on as a healthy company," O'Leary said. Companies kept alive this way are starved of the capital they'd need to invest and grow, which he said will hurt eventual recovery values
Panossian put a number on the vintage's share of current defaults. "In the 2025 2026 time frame, the defaults that we saw were 74% from this vintage" — loans issued in 2021 or 2022 that carry 7-year maturities landing in 2028 or 2029, mostly in a broadly syndicated loan market with few covenants, so they fail on cash flow rather than a covenant trip
He expects the maturity wall to get worse, not better, as it approaches. These were often asset-light businesses starved of cash for years, and Panossian said that when they finally default, loss severities will be very deep and driven by cash flow and maturity rather than choice. He called the coming defaults a "need to have restructuring" rather than a discretionary one
Any company that could refinance already has. "But what I would say is any good CFO in the market right now has already refinanced that vintage of loan. If they still have their 2021 or 2022 era issuance loan outstanding in 28 or 29, it means they can't refi generally."
4. Software's AI Discount
Panossian gave the fullest account of software's exposure across the credit markets and why he thinks the sector's assumed diversification stopped working after ChatGPT's release in November 2022.
Exposure varies enormously by market. Software is 3-5% of high yield, 12-15% of broadly syndicated loans (the number-two sector after healthcare services), and anywhere from 9-10% to 70% of a given BDC's book in direct lending
The old diversification argument was that software wasn't really one industry — you looked through to end markets. Healthcare software, government services, insurance software and mining-exploration software all behaved differently. "Now after November of 2022 when ChatGPT was released... the perception that software outcomes can be correlated can become correlated because of AI disruption is a new consideration"
Five factors, in Panossian's framework, separate a durable software moat from a vulnerable one: proprietary data no model can train on, regulatory lock-in, network effects, transaction embedding, and being the system of record. "We focus on software that is deeply embedded in the daily workflows and business processes of customers. They require meaningful buyin from their stakeholders both internal and external. Very high switching costs."
Even a durable moat doesn't guarantee a refinancing. "I don't think you can rely even when you identify those factors on the market necessarily refinancing out that credit. You have to rely on what you can get out of the cash flows," Panossian said
The leverage math on these companies has gone from comfortable to dangerous. Software companies bought in 2021 or 2022 at 10 to 14 times revenue and levered three to five times revenue have seen valuation multiples fall 60% to 70% since. Many paid non-cash, payment-in-kind interest for years rather than cash, adding to the loan balance the whole time — so a company that started around 35% loan-to-value in 2021 or 2022 can be near 100% loan-to-value by 2028 or 2029
The same private equity sponsors also cut the wrong staff. To flip from PIK to cash-pay coupons, sponsors "fired their salespeople, their R&D staff. They ran it for cash flow" — "exactly the wrong positioning" when competing against AI requires more R&D, not less. Panossian said a company would need to show it can actually compete, not just hold a defensible moat, before Oaktree gets excited about it again
5. The Rotation to Hard Assets
Whitelaw asked whether software's troubles were pushing capital toward heavier, harder-to-disrupt assets.
Panossian said the acronym investors were using a year ago — HALO — has become a return to liking hard-asset-heavy businesses, with renewed interest in asset-backed finance, where loans and leases are secured by physical collateral, and a resurgence in infrastructure lending against bridges, airport concessions and other collectible assets. "Again, bridges, airport concessions, rare and collectible assets that are not easily printed through a computer program."
6. Financing the AI Buildout
Whitelaw's final prepared topic was AI capex, where Panossian said credit has to play a much bigger role than it has historically.
The buildout needs an enormous amount of capital. "It's 5.5 trillion. I've seen numbers as high as 7 trillion over the next 5 to 8 years." Between structured products and high-grade leveraged finance, credit markets face close to $3 trillion of funding need "that has been created almost overnight" — a financing gap that barely existed as a topic four years ago
Some of the structures carry a guarantee gap. Special-purpose vehicles built around hyperscaler lease payments don't always carry a full guarantee from an investment-grade counterparty, which leaves lenders exposed if technology changes, construction or permitting slows, or power access falls short
Correlated risk sits inside the complex, not just across it. A change in chip quality, power needs per chip, large-language-model efficiency, or power technology could all move together across the data-center sector, raising the question of what a lender faces at refinancing
Spreads have started widening in recent weeks, which Panossian reads as either saturation or a market pricing in these risks. He framed direct lending's future returns as split between a retrospective factor — the software exposure already locked into portfolios — and a prospective one, AI data centers, where the outcome is still being written
O'Leary agreed and flagged the next inflection point. "You're approaching saturation in certain key parts of the market spreads are starting to widen out and it may be at some point that you have to go from investment grade to other pockets of capital. You'll see a step function jump in that cost of capital when that happens."
7. Credit Secondaries Widen
Returning to the liquidity theme, O'Leary said the combination of BDC redemptions and stretched fund lives is pushing more activity into the credit secondary market, historically dominated by private equity stakes rather than credit ones.
LPs were promised seven-year fund lives and are being asked to wait longer. "We're now stretching beyond that. And I think LPs are getting anxious for two reasons. One, they need the capital, pure and simple," O'Leary said, and two, a fund that won't return capital makes an LP question whether they can trust the manager's marks
LPs are responding by selling. "LPs are bundling up their stakes in private credit funds and putting them out the market. We've seen two or three very large portfolios come to market just recently," O'Leary said, and while they haven't traded at large discounts to face value yet, he expects discounts to widen as volume picks up — which he called a real opportunity for Oaktree
Oaktree's underwriting edge in secondaries comes from having priced many of these same credits at issuance, even in cases where Oaktree didn't originally participate, layered with an understanding of what's happened to the business since — "in a lot of cases again there are things that are cropping up that were never really anticipated two or three years back. Software being the biggest example," O'Leary said
8. Where Else They're Lending
Whitelaw's audience-submitted question asked what O'Leary finds attractive outside of corporate credit.
Specialized, hard-to-source lending is the common thread. Life sciences lending requires understanding both the science and the structure of a borrower, which limits competition and keeps the opportunity set attractive through cycles
Asset-backed finance has a structural tailwind: regional banks pulled back sharply after the collapse of Silicon Valley Bank, leaving an abundance of deal flow without enough capital supply to meet it. O'Leary said the work here is as much about underwriting the people originating the debt as the assets themselves. "You have to be able to sniff out the cockroaches and the fraud."
Infrastructure lending, with Brookfield as a partner, benefits from an information advantage in a market few firms understand globally, spanning data-center lending and other large-scale projects that need financing capital isn't readily available for at that scale
9. Marking Down BDC Assets
Whitelaw closed by asking Panossian whether BDC managers are marking their portfolios down enough, and by what process.
There's no standard playbook. "There isn't a manual that says a BDC must have X, Y, and Z as part of its process," Panossian said — it's largely a judgment call blessed by each fund's board, and it varies from manager to manager
Oaktree's own rule is to mark down for either of two reasons: fundamental performance degradation that implicates default or loss risk, or a market-wide widening of spreads that would force a lower sale price even for a performing company. Positions aren't meant to be marked to market, but Oaktree overlays its own view of where spreads have moved
Panossian said he can't speak for every manager, and some have been erratic. "I can't say the same for every BDC manager and we have seen unfortunately some BDC managers take a position that would par one quarter down to zero the next quarter and that has created some sensitivity and some skepticism on the part of investors as to whether BDC's are appropriately marking their books or not."
Bonus Insights
Panossian's own playbook for now is patience over leverage. "What we've been doing is what we will continue to do, which is to focus on bottoms up credit fundamentals to inform a patient deployment of capital and not an aggressive use of leverage to generate higher dividends in our vehicles," he said, adding that Oaktree wants to hold dry powder for "those periods of time where there's pockets of weakness"
O'Leary said nothing in the macro data points to an imminent credit cycle. He framed the $250 billion distressed-debt universe as largely a function of higher rates rather than stress already spreading through the broader economy
O'Leary and Panossian's bottom line is that private credit's current troubles aren't confined to a handful of bad borrowers: they trace to a specific 2021-2022 vintage still working through its maturity wall, to a software sector whose diversification broke down once AI made its end markets correlated, and to a similar leverage pattern already visible in the loans financing the AI buildout that's supposed to be the market's next growth story.
Books & Resources Mentioned
Sea Change – Howard Marks (The December 2022 Oaktree memo O'Leary cites as the source of the major themes — rising rates, consumer divergence, dispersion — still driving credit markets today)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: