Key Takeaways
- • Carlyle executives Jason Thomas and Mark Jenkins warn in a whitepaper, "Financing AI Without a SaaS Repeat," that direct lenders risk repeating the concentration errors that underwrote the software-as-a-service boom as they finance AI infrastructure.
- • PitchBook LCD data shows nearly 10% of total debt investments held by the largest BDCs were tied to borrowers with at least one distressed position as of Q2 2026.
- • Distressed debt at the major BDCs rose 64% year-over-year to $6.8 billion, or 5.2% of BDC debt assets, with software loans and second-lien tranches accounting for 89% of the increase.
- • Analysts at Morgan Stanley and JPMorgan estimate the AI compute buildout will require $1.5 trillion to $3.2 trillion in capital, with private credit expected to absorb over $1 trillion, roughly 40% to 60% of global private credit AUM depending on the benchmark.
- • The author argues lenders should institute explicit, firm-wide concentration limits, drawing on OCC "Concentrations of Credit" guidance, before the buildout peaks.
Summary
Carlyle executives Jason Thomas and Mark Jenkins, in a whitepaper released this week titled "Financing AI Without a SaaS Repeat," caution direct lenders against repeating the structural concentration errors that underpinned the software-as-a-service boom. As private credit managers deploy capital into the roughly $1 trillion artificial intelligence infrastructure buildout, the authors argue that lenders should apply serious credit due diligence to AI-related loan portfolios, particularly concentration risk. The Carlyle analysis notes that underwriting standards shape which transactions private equity sponsors pursue: relaxed credit boxes enabled sponsors to bid higher valuation multiples and target more software acquisitions than they otherwise could have.
Sycamore Tree Capital Partners made a similar point earlier in the summer, observing that AI infrastructure financing has migrated from hyperscaler balance sheets into external credit markets, with risk increasingly sitting in less visible and less liquid positions. According to the Carlyle authors, the structural warnings are already visible in market data. A PitchBook LCD analysis by Sebastian Kian, titled "Nearly 10% of top BDCs' investments tied to borrowers with a distressed position," found that as of Q2 2026 nearly 10% of total debt investments held by the largest business development companies were tied to borrowers with at least one distressed debt position. Distressed debt, defined as positions with a fair value-to-cost ratio below 80%, rose 64% year-over-year to $6.8 billion, representing 5.2% of BDC debt assets. Software loans and second-lien tranches accounted for 89% of that increase.
The whitepaper contends that moving from SaaS into AI compute infrastructure amplifies rather than escapes concentration risk. Three factors are cited: uncertainty over how long-term profits will be divided among chip makers, foundation model developers, infrastructure owners and application software providers; accelerated depreciation of GPU-based compute assets as hardware iterates rapidly, which could render multi-year debt assumptions obsolete; and macroeconomic interdependence, since AI capital expenditure now accounts for a large share of economic growth, meaning AI credit defaults could correlate with broader downturns.
The author draws on bank regulatory practice, noting that the Federal Reserve, OCC, FDIC and state regulators have long treated sector concentration as a serious failing. The OCC's "Concentrations of Credit" booklet defines a credit concentration as any pool of loans sharing common sensitivities to financial or technological developments, warns that concentrations can become problematic even without past problems, and calls for quantitative limits, portfolio-wide stress testing and contingency plans. Although bank regulators do not supervise unlisted private credit funds, the author argues those principles offer a useful blueprint.
Analysts at Morgan Stanley and JPMorgan estimate the AI compute buildout will require $1.5 trillion to $3.2 trillion in capital, with private credit expected to absorb over $1 trillion, equal to about 40% to 60% of global private credit AUM depending on the benchmark used. Carlyle's message is not that lenders should avoid AI infrastructure, which may offer complexity premiums across data center construction, power financing and structured equipment leases, but that alternative asset managers should set firm-wide concentration limits before the buildout peaks.
Sycamore Tree Capital Partners made a similar point earlier in the summer, observing that AI infrastructure financing has migrated from hyperscaler balance sheets into external credit markets, with risk increasingly sitting in less visible and less liquid positions. According to the Carlyle authors, the structural warnings are already visible in market data. A PitchBook LCD analysis by Sebastian Kian, titled "Nearly 10% of top BDCs' investments tied to borrowers with a distressed position," found that as of Q2 2026 nearly 10% of total debt investments held by the largest business development companies were tied to borrowers with at least one distressed debt position. Distressed debt, defined as positions with a fair value-to-cost ratio below 80%, rose 64% year-over-year to $6.8 billion, representing 5.2% of BDC debt assets. Software loans and second-lien tranches accounted for 89% of that increase.
The whitepaper contends that moving from SaaS into AI compute infrastructure amplifies rather than escapes concentration risk. Three factors are cited: uncertainty over how long-term profits will be divided among chip makers, foundation model developers, infrastructure owners and application software providers; accelerated depreciation of GPU-based compute assets as hardware iterates rapidly, which could render multi-year debt assumptions obsolete; and macroeconomic interdependence, since AI capital expenditure now accounts for a large share of economic growth, meaning AI credit defaults could correlate with broader downturns.
The author draws on bank regulatory practice, noting that the Federal Reserve, OCC, FDIC and state regulators have long treated sector concentration as a serious failing. The OCC's "Concentrations of Credit" booklet defines a credit concentration as any pool of loans sharing common sensitivities to financial or technological developments, warns that concentrations can become problematic even without past problems, and calls for quantitative limits, portfolio-wide stress testing and contingency plans. Although bank regulators do not supervise unlisted private credit funds, the author argues those principles offer a useful blueprint.
Analysts at Morgan Stanley and JPMorgan estimate the AI compute buildout will require $1.5 trillion to $3.2 trillion in capital, with private credit expected to absorb over $1 trillion, equal to about 40% to 60% of global private credit AUM depending on the benchmark used. Carlyle's message is not that lenders should avoid AI infrastructure, which may offer complexity premiums across data center construction, power financing and structured equipment leases, but that alternative asset managers should set firm-wide concentration limits before the buildout peaks.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.