The CEO of a $23 billion asset manager says private credit is dangerously overexposed to software companies leveraged 10x, and AI is about to test every thesis that funded them.

Marathon Asset Management CEO Bruce Richards told Bloomberg Television on February 26 that default rates in software-exposed private credit could exceed 15%, calling the industry's concentration in software lending a "train coming down the tracks that you could see from some distance."¹

Richards said the overall direct lending default rate could reach 5% to 6% over the next few years, but warned it will be dominated by software sector losses. He called lending to software companies at current leverage levels "not intelligent," noting many borrowers carry approximately 10x leverage in private credit portfolios.²

Key facts:

_{Sources
1 Bloomberg | Marathon's Richards Fears 15% Direct Loan Software Defaults
2 Seeking Alpha | Private Credit Default Risk in Software Could Top 15%
3 Prime Buchholz | Software Stress and AI Risk in Private Credit
4 24/7 Wall St | How BDC ETFs Are at Risk Under Private Credit Default Crisis
5 Yahoo Finance / UBS | Private Credit Exposure to AI Disruption High, Not Priced In
6 CNBC | Private Credit Worries Resurface as AI Pressures Software Firms
7 CNBC | Blue Owl's Software Lending Triggers Another Quake in Private Credit
8 Bloomberg | Private Credit Market Split on Crisis Risk as Software Defaults Rise
9 With Intelligence | Private Credit Outlook 2026: The Market Faces Its First Big Test
10 Morgan Stanley | Private Credit 2026 Outlook
11 BlackRock | Putting Private Credit Concerns in Perspective
12 PineBridge | Private Credit Direct Lending Outlook: Tailwinds for 2026
13 ClassAction.org | Figure Lending Facing Class Action Over Data Breach
14 TechCrunch | Figure Confirms Data Breach
15 NY Governor | Hochul Announces BNPL Regulation
16 Regulatory Oversight | NY BNPL Licensing Regime
17 Consumer Finance Monitor | Oregon HB 4116 DIDMCA Opt-Out
18 JD Supra | Oregon DIDMCA Opt-Out
19 Auto Finance News | Yendo Triples Volumes After $200M Injection
20 Ventureburn | Yendo Secures $200M
21 GlobeNewsWire | authID Selected by POS Lender
22 Stock Titan | authID Biometric Fraud Fight}

Private credit grew from roughly $800 billion in 2020 to over $3 trillion by 2025. A disproportionate share of that growth flowed into software companies. The thesis was simple: SaaS businesses generate predictable recurring revenue, have high gross margins, and benefit from customer switching costs that make cash flows durable. Private credit funds lent against those cash flows at 5x to 10x leverage, pricing in the assumption that the revenue would keep recurring.³

AI is now challenging that assumption. As Richards wrote on X following his Bloomberg interview: "While there will be many great equity investments to be made in software/AI, expect default rates to increase dramatically as AI will lead to creative destruction (positive) and destruction (negative) of software businesses."¹

The concern is specific. AI tools can replicate functions that previously required dedicated SaaS subscriptions. When a $200/month/seat SaaS product can be replaced by an AI agent at a fraction of the cost, the recurring revenue that justified the lending thesis evaporates. The borrower still owes 10x leverage against cash flows that are shrinking, not growing.

The numbers are substantial. Software and technology companies represent approximately 25% of the private credit market by value.³ In BDC portfolios specifically, software accounts for 17% of deals by count, the second-largest sector after commercial services. Roughly $40 billion, about 10% of total BDC assets, overlaps with public loan markets, with the heaviest concentration in tech.⁴

The stress signals are already visible. Covenant defaults reached 3.2% as of September 2025, well above the 1-2% payment default range that has held since 2021.³ Software companies account for the largest share of payment-in-kind (PIK) loans, where borrowers delay paying interest in cash. PIK arrangements are standard for fast-growing companies building revenue, but they become dangerous when a borrower's fundamentals weaken: deferred interest compounds into a credit problem quickly.

Blue Owl Capital's software lending portfolio has triggered specific investor concerns, with CNBC reporting that illiquid loans and investor redemption pressures created "another quake in private credit" on February 20.⁷ And UBS published research warning that in an aggressive AI disruption scenario, U.S. private credit defaults could climb to 13%, significantly higher than the stress projected for leveraged loans (8%) or high-yield bonds (4%).⁵

Because the capital that funds your warehouse lines, your credit facilities, and your investor commitments comes from the same pool. Private credit is interconnected. When a BDC takes losses on software loans, it does not selectively reduce its software exposure. It tightens credit across its entire portfolio to shore up capital ratios. If your warehouse lender or credit facility provider has significant software exposure, their stress becomes your constraint.

The maturity wall makes the timing urgent. Twenty-three out of 32 rated BDCs have unsecured debt maturing in 2026, totaling $12.7 billion, a 73% increase over 2025.⁴ BDCs that need to refinance maturing debt while absorbing software-sector losses will face tighter terms, higher costs, or both. That pressure rolls downhill to every borrower in their portfolio.

Richards offered a partial reassurance: he sees limited contagion risk to the broader market from software defaults. But "limited contagion" at the macro level does not mean "no impact" at the individual facility level. If your capital partner has 20% of its book in software and that segment defaults at 15%, the losses affect the partner's overall portfolio economics, including the terms it offers you.

Map your capital providers' software exposure. Ask your warehouse lender, your BDC investor, or your credit facility provider what percentage of their portfolio is in software and technology. If they will not tell you, that is an answer. If the number exceeds 20%, stress-test what happens to your facility terms if their software book deteriorates.

Diversify your capital sources. If your lending operation depends on a single warehouse facility from a BDC with heavy software exposure, you have concentration risk. The time to secure backup facilities or alternative capital partners is before the stress hits, not during it.

Watch for tightening signals. Credit facility providers do not announce portfolio stress. They signal it through tighter advance rates, higher interest spreads, increased reporting requirements, or reduced commitment amounts at renewal. If your next facility renewal comes with materially different terms, software-sector losses may be the unstated reason.

Most lenders fight regulation like it's an obstacle to profitability. But regulation exists because there was a problem that needed solving.

The winners ask a different question: How can we use these regulations to build a better company?

Better transparency leads to better customer relationships. Better compliance frameworks create opportunities for repeat business.

The pendulum might swing too far initially, but it always settles in the middle. Smart lenders position themselves there first.

Listen to the full podcast episode

Figure Technology Solutions (FIGR), the largest nonbank HELOC lender in the U.S., confirmed a data breach on February 25 after ShinyHunters published 2.5GB of stolen borrower data on the dark web. The attack exploited a single employee via voice phishing to obtain Okta SSO credentials and MFA codes. Nearly 967,200 accounts were compromised. A class action (Mardikian v. Figure Lending, 3:26-cv-00135) alleges GLBA Safeguards Rule violations and delayed breach disclosure. For every lender using cloud-based SSO: if your MFA can be read over a phone call, you are vulnerable to the same attack.^{13 14}

Governor Hochul and the NY DFS published proposed rules on February 23 requiring licensing, supervision, and consumer protections for all buy-now-pay-later providers in New York. The framework caps interest-bearing BNPL loans at 16%, mandates risk-based underwriting, imposes fee limits, and establishes dispute resolution. The move fills a regulatory vacuum created when the CFPB reversed course on classifying BNPL as credit. Pre-proposal comments close March 5. California, Illinois, and Massachusetts are expected to follow.^{15 16}

Oregon's House Committee approved HB 4116, which would prohibit out-of-state banks from making consumer loans of $50,000 or less to Oregon borrowers at rates exceeding 36%. The bill failed in 2025 but gained momentum after the Tenth Circuit's ruling in NAIBankers v. Weiser upheld Colorado's ability to enforce state usury limits on out-of-state bank loans. For alternative lenders using bank partnership models to export rates above state caps, this directly threatens program economics.^{17 18}

Dallas-based Yendo, creator of the first vehicle-secured credit card, secured a $200M warehouse facility from returning investor i80 Group and plans to triple application and origination volumes in 2026. Consumers across 45 states access vehicle equity as collateral for up to $10,000 in revolving credit at prime rates, bypassing credit scores. AI-powered underwriting verifies vehicle collateral in minutes. The company grew 700% after Series A and has saved customers over $150M in interest versus alternative lenders.^{19 20}

A U.S. point-of-sale lending platform selected authID (AUID) for biometric identity verification across merchant onboarding, consumer origination, and account access. The system replaces risk scores and SMS-based MFA with deterministic biometric proof at a 1-in-1-billion false positive rate. With U.S. payments fraud losses at $12.5 billion in 2024 and synthetic identities driving 45% of auto lending fraud, the deployment signals growing demand for identity-based verification in consumer finance.^{21 22}