AI Jitters, Private Liquidity Crunch and 26% BDC Discounts: Why Two Pros Still See Opportunity

Click Here to Download Article

AICA’s Quarterly BDC Earnings Pulse

March 17, 2026

The latest selloff in business development companies (BDCs) has been framed as an “AI SaaS apocalypse” colliding with a looming liquidity crunch in private credit, but two of the sector’s most closely watched practitioners see more cyclicality than catastrophe in the headlines.

Speaking on the Active Investment Company Alliance’s March 11th BDC Earnings Pulse call, Putnam Investments’ Mike Petro and Wells Fargo’s Finian O’Shea argued that disciplined underwriting, structure and permanent capital leave listed BDCs better positioned than the tape suggests—even as software non‑accruals rise, spreads widen and non‑traded vehicles confront a B‑REIT‑style stress test.

AICA’s BDC “pulse check”

The BDC Earnings Pulse is AICA’s fast‑paced, quarterly debrief designed to connect leading BDC analysts and institutional investors with the latest earnings trends and forward positioning. Moderated by

AICA Executive Chairman and CEF Advisors President John Cole Scott, the series aims to move beyond raw numbers to assess credit quality, NAV performance, dividend coverage and how portfolios are being positioned for the next three to six months across the listed BDC landscape.

March’s edition, “BDC Earnings Pulse: Trends, Takeaways & Tactical Positioning,” centered on two themes dominating investor conversations: AI‑linked software risk and the implications of rising redemption pressure innon‑traded and interval BDC structures.

Scott opened by quantifying the damage. CEFData’s equal‑weight BDC index is down roughly 13% in total return since January 15 and about 20% over the past year from last February’s levels, a drawdown he noted has been widely blamed on software and AI exposure inside BDC portfolios.

“From our research…about 25% of the holdings appear to be software, even though not always disclosed by the BDCs themselves,” he said, estimating that across the universe CEFData tracks, roughly 1,150 portfolio companies are software focused. With that backdrop, he asked Petro to address the alarmist “AI SaaS apocalypse” narrative that has taken hold in parts of the market.

AI, software and the “SaaS apocalypse”

Petro, who manages Putnam’s BDC income and small‑cap value strategies, said the core fear is straightforward. “People are worried that AI is going to make many of these software companies’ platforms, or at least their business model, obsolete,” he explained. The bear case is that AI agents will learn to interact directly with corporate databases and workflows, disintermediating existing software platforms and undermining the traditional seat‑based user model that underpins many recurring revenue credits.

Exposure within BDCs is material but highly uneven, ranging from the low single digits at some lenders to roughly 70% at vehicles like Blue Owl Technology Finance, which he described as “all software, all the time.”

Still, Petro is skeptical that AI will turn software books into a wholesale credit disaster. “Like a lot of these themes that Wall Street latches on to, I think that there’s some truth to the matter, but there’s also a whole lot of speculation and exaggeration too,” he said. Historically, software assets in BDC portfolios have generated lower loss rates than the broader universe, with Scott citing CEFData’s preliminary estimate of roughly 50 basis points of non‑accruals in software versus about 135 basis points across all sectors.

The notion that those losses will suddenly leap to “four or five or 15%” in the near term is not something he “just sees coming.”

For Petro, the key differentiator will not be abstract AI theorizing but fundamental credit work tailored to software. “Overall credit underwriting skill is number one,” he said. “I would rather have a manager that has software exposure but a very good credit underwriter than a poor credit underwriter with no software exposure.” He pointed to covenant packages that tie directly into software key performance indicators as a critical early‑warning tool.

“You want there to be coverage around key performance indicators or KPIs that matter for software—an important one would be retention of customers,” he said. If AI or competitive pressure starts to erode a borrower’s model, it should surface in renewal pricing and volumes, triggering covenant tests and giving lenders room to enforce changes—”maybe sell the company, make some sort of change while there’s still time to recover some capital.”

He also urged investors to distinguish between business models and borrower profiles. BDCs with long experience in the space, he said, “have been moving towards those companies that have more transaction‑based and more value‑based pricing” rather than pure seat‑based licenses that are directly tied to headcount and potentially more vulnerable as AI reshapes white‑collar work.

Company size matters as well. “The software companies that are smaller as a group are probably going to have a little more—or a lot more—risk than the larger ones,” Petro argued. At scale, vendors tend to be more deeply embedded in customer processes and data, can charge more for that value and have the engineering bench to integrate AI features themselves, giving them more defenses and “stickiness” against disruption.

To reflect that dispersion, he believes lenders should insist on lower loan‑to‑value ratios in the most exposed tails: “There are going to be some losers in this AI battle, and the more headroom you have in your capital stack above you, the better chance of actually getting some of your capital back.”

From pristine to normal: software’s new status

O’Shea, who covers a broad universe of publicly traded BDCs at Wells Fargo, largely agreed with Petro’s contours but framed AI as an accelerant of trends already underway. “I think what it might do is accelerate pressures that were maybe already happening, and that’s not too bad of a thing,” he said. In his view, software has quietly migrated over the past few years from a sacrosanct bucket inside BDC portfolios to one that behaves more like “normal” corporate credit. “Software went from sort of a pristine part of the portfolio to something that’s looking more and more normal,” he observed. “There are actually software non‑accruals nowadays, which is kind of a new thing.”

Competition, pricing pressure and lower switching costs are all part of that normalization, and AI may amplify each of them at the margin. “Maybe AI accentuates that from a switching‑cost perspective, or a seat‑based perspective, as Mike hit on,” O’Shea said. As long as a company’s value proposition is genuine rather than the product of a transient “sales push,” however, he believes many credits will navigate the transition.

The market, he suggested, has already priced in an uptick in software loss rates, with that bucket now likely to see “a little elevated” losses, “say, in line or above the rest of the industry,” which “probably contributes a little bit to today’s discounts.”

Scott pressed the panel on whether, by the same logic, software should also be the first part of the portfolio to benefit from AI. He pointed to customer‑experience platforms like Zendesk as emblematic of that dual‑edged dynamic, where AI tools can both commoditize incumbent offerings and turbocharge productivity. O’Shea used the prompt to restate a basic credit‑risk principle. “Just to state the obvious, in a credit book, you don’t want eight of your portfolio companies to double in EBITDA and two of them go to zero,” he said. Even if AI creates big equity winners, the lender’s nightmare is a barbell of home runs and zeros, not a smooth curve of modest outcomes.

The credit cycle context: fraud, losses and experience

Beyond AI, both Petro and O’Shea see more traditional late‑cycle concerns brewing. “There is certainly worry in the credit market,” Petro said, but what’s unusual is that “it’s not really about the economy, which, by the way, is what really drives credit most of the time.” Instead, he pointed to three drivers: AI’s impact on software, several recent high‑profile borrower frauds and a generalized sense that “there’s too much money that’s flowed into the space chasing” higher‑yielding private loans.

Specifically, he drew a sharp line between newer private pools and the listed BDC cohort. “The exposure of the public BDCs to these frauds has been super low,” he said, adding that in the managers he considers the strongest the impact has been “zero,” and some even reported they “could have invested in these deals, but we saw red flags that kept us out.”

Loss data supports a more tempered outlook. Petro cited roughly 65 basis points of annualized losses as a reasonable baseline in his own experience and compared that with the industry’s performance during the global financial crisis. “The BDCs…were, I think everyone would agree, not nearly as capable during the great financial crisis—the cumulative losses during that were less than 15%, and that was an actual financial crisis,” he said. With today’s platforms “much bigger, much better, much more sophisticated, better asset quality, better liability quality,” he does not see a path to catastrophic double‑digit loss outcomes driven by AI or otherwise.

Private BDCs, interval funds and redemption stress

If AI and software dominate the narrative, liquidity pressure in non‑traded and interval vehicles may be the more immediate mechanical risk—and opportunity—for listed BDCs. Scott noted that the tender deadline for the Cliffwater Private Credit interval fund had just passed, amid talk that tender requests could exceed the 7% quarterly cap he encourages interval managers to target if ever needed. He also highlighted how quickly the private universe has grown: “There’s 120+ non‑listed private BDCs now in our database, and that was probably 30 ten years ago,” he said.

O’Shea called the non‑traded side “where the sort of suffering is right now, or at least the headlines are.” “It’s both, right—it’s headlines and headline‑driven suffering,” he said, arguing that the harsh tone has “probably spilled into public BDCs’ share‑price pressures,” which he views as “pretty undue.”

Provided the macro backdrop avoids a deep recession, he sees the structural stress as ultimately positive for listed vehicles. A B‑REIT‑style cycle of tender caps, redemption gates and occasional portfolio sales should “move the market in a better spread‑widening direction,” he argued, allowing BDCs to originate or refinance loans at higher yields than they have seen in years.

For now, he expects managers of the largest non‑traded platforms to “tap the brakes on new deployment” while they assess the durability of outflows. Liquidity and credit profiles remain solid, and the episode could resolve quickly. “I could see this all go away next quarter. I could see it last a few more quarters,” O’Shea said, noting the presence of “a good…institutional wall of money” behind portfolio sales completed so far.

The permanent capital advantage: arbitrage opportunities emerge

Petro extended the analysis to capital formation and check sizes across multi‑product platforms that straddle listed and non‑traded vehicles. “You’re starting to see already some of these platforms that have a lot of exposure to the mass‑wealth channel that are seeing these redemptions in their private vehicles,” he said.

Even if they also manage public BDCs, “those publics—or the whole platform rather—is going to have to make some slight degradation in their check size,” especially if redemptions continue and non‑traded funds impose BREIT or Starwood‑style gates for extended periods.

That shift, he argued, sets up a relative advantage for vehicles with truly permanent capital and no quarterly redemption feature. “Who’s going to benefit? The ones that don’t have that source of redemption, who are less exposed to mass wealth, the ones who have more permanent capital,” Petro said. Under‑levered listed BDCs such as Blue Owl Technology Finance—at roughly 0.75x debt‑to‑equity—and Kayne Anderson’s KBDC, closer to 1.0x, can “lean in,” either by adding assets or by rolling run‑off into wider‑spread paper.

He also highlighted a simple arbitrage that some allocators are already discussing: “You could sell a private BDC at 100% of NAV, and then you trade into a very similar one at 75 or 80% of NAV—oftentimes the exact same manager and a portfolio that looks a hell of a lot like the one you’d be selling in private.”

Audience questions focused on whether interval funds facing high tenders might be forced to dump loans at discounts, thereby resetting marks across overlapping BDC portfolios. O’Shea doubted that well‑run platforms would go down that path. “Managers will diligently only sell stuff at par, or at least fair‑market value, that they already have,” he said. Before any manager sells performing assets at a “NAV‑impairing discount,” he expects them to fully use pro‑rata tender mechanics and portfolio cash flows. “Selling performing assets at a deep discount would be hard to come back from for a non‑traded BDC.” I could be described as an action that could have reputational implications for retail shareholders.

Petro emphasized that the structures were built precisely to avoid forced asset sales. “First of all, these loans don’t really trade—that’s the whole point of private lending,” he said, noting that reliable bids can take weeks to assemble and often carry a liquidity discount. That illiquidity is why investors accepted quarterly gates and caps in the first place.

“On average, your duration is about five years—that’s 20% a year your portfolio turns over, that’s 5% a quarter, which is exactly what they set the liquidity gate at,” he explained. While turnover varies, the math supports gate levels, and boards are likely to “enforce them” rather than “disadvantage their shareholders who are sticking with them” by selling loans into weak bids.

Scott added a structural nuance that matters for advisors. Interval funds must operate within prospectus‑set repurchase ranges and cannot simply raise liquidity above thresholds—”if it’s stated 5% like the Cliffwater fund, they cannot go over 7%, which is…because that’s the rule,” he said—whereas boards of non‑traded BDCs running tender‑offer programs have more discretion to increase or decrease liquidity. That makes client education critical, particularly because many brokerage statements still code interval funds as if they were open‑end mutual funds.

Buybacks, fees, insider signals and the outlook

With discounts wide—Scott noted that an RSI‑style study of CEFData’s equal‑weight index shows only about 3% of observations have seen downside volatility as acute as late February’s reading—participants pressed the panel on board‑level responses. O’Shea said he expects to see “a little bit of buybacks” but “not too much on a sustainable basis.”

BDC managers may be somewhat reluctant to revisit broad fee‑reduction discussions, given prior changes and the relatively strong trading levels seen until recently.

Finian added that he expects the current loss‑rate trajectory—roughly double to triple his 65‑basis‑point baseline in some cases—to continue through this year as Covid‑era vintages and pre‑rate‑shock deals are worked out, with a gradual normalization beyond that if AI and macro risks do not materially deteriorate.

Petro sees the first quarter 2026 (next earnings season) as a more revealing test of board behavior than the fourth quarter, given how far valuations have reset. With the “AI‑software panic” having driven some names down “quite dramatically,” he senses many boards believe discounts are “overdone” and want to “react and take advantage” through repurchases. At the same time, he is already hearing the familiar argument that the best use of marginal capital is to “invest in these improving spreads”—a perennial competitor to buybacks when front‑book returns move higher.

On alignment, Petro pointed to the growing use of total‑return lookbacks that claw back incentive fees if a BDC fails to meet a return‑on‑equity hurdle over a multi‑year period. “If a BDC doesn’t achieve a certain ROE over, let’s say, a three‑year trailing period, then they have to make up the difference by cutting their incentive fee,” he explained. While that will not prevent share prices from reacting to credit hits, it “does preserve the dividend a little longer than it would otherwise, because the management has to eat that shortfall.” He called lookbacks a “good signaling device for management alignment” that concentrates minds in a sector where external management structures can create tension between growth and shareholder returns.

Insider buying drew a more nuanced response in the session, with the focus on both the pattern of activity and who is doing the buying. The discussion emphasized that the most compelling signal often comes when mid‑level employees commit a meaningful portion of their own compensation to shares, whereas large purchases by the C-Suite tend to be interpreted as more symbolic, closer to a modest buyback than a high‑conviction bet.

Participants also noted that, while there is usually at least some insider activity in periods like this, the overall level has been fairly limited, and many believe a more robust wave of buying would be a meaningful support for investor sentiment.

As the session closed, Scott asked for a directional view on sector valuations heading into year‑end, anchoring his question on CEFData’s volatility work and the current 0.74x median price‑to‑NAV multiple. Petro, who “buys and sells these all the time,” agreed that discounts look too wide. “The median BDC is trading at 0.74 times [NAV] by December 31, he expects that figure closer to 0.80x, implying roughly 500 basis points of valuation upside on price‑to‑book alone before factoring in cash yields. He does not expect a quick return to 0.85–0.90x multiples because key fears—AI’s long‑term impact on software business models and the true elasticity of private‑credit capital under stress—”can only be resolved over time.”

In the meantime, the panelists see a familiar late‑cycle pattern: slowly rising losses, tighter credit, more dispersion and opportunities for well‑situated, permanent‑capital vehicles to lean into wider spreads. “Credit is going to get worse, and some of the pundits are going to be like, ‘Aha, I told you, the end is nigh,'” Petro said. “But why are we getting worse? Well, because that’s where we are in the credit cycle.”

For investors who can distinguish between AI rhetoric, liquidity optics and actual underwriting, he and O’Shea suggest that today’s combination of spreads and discounts looks more like a tactical opportunity than a systemic threat.