AI debt frenzy sparks liquidity fears in private credit

AI-driven disruptions have exacerbated valuation and liquidity challenges in private credit markets, particularly for software loans. Traditional liquidity mechanisms like CLOs function well for conforming loans but falter with software and tech loans, which lack standardized covenants and predictable cash flows, complicating both valuation and secondary market trading. As one analyst noted, the private credit secondary market remains underdeveloped for non-conforming loans to cash flow-negative tech companies, with total volume around $10 billion—minuscule compared to the broader private credit market. This liquidity squeeze has prompted increased redemptions driven more by investor caution than clear evidence of loan underperformance, intensifying pressure on fund managers to address these structural mismatches.

Liquidity management in private credit funds is under strain as redemption requests frequently exceed designed limits, with some Business Development Companies (BDCs) facing quarterly redemption demands up to 40%, far above their typical 5% allowance. This mismatch creates a coordination problem where investors race to exit, even when credit quality remains intact, because redemptions are often funded through asset sales rather than cash flows. JPMorgan’s recent tightening of lending standards amid software loan repricing underscores the growing caution in underwriting, reflecting the sector’s vulnerability to AI-driven market shifts. Meanwhile, the opacity of fund structures—evidenced by increased use of payment-in-kind interest and off-balance-sheet borrowing—further clouds true leverage and liquidity risks, complicating accurate valuation and risk assessment.

The traditional valuation paradigms for software loans are being upended by AI’s impact on underlying business models and collateral quality. Historically, lenders valued software loans based on high gross margins, sticky enterprise customers, and the ability to flex costs to enhance free cash flow, often lending against recurring revenue multiples as a proxy for EBITDA multiples. However, AI introduces new uncertainties about the durability of recurring revenues and enterprise value as collateral, prompting lenders to reassess these assumptions. As one expert put it, "What is the collateral you have in a software borrower? Maybe it's recurring revenues, maybe it's enterprise value," highlighting the fundamental rethinking underway in loan underwriting and risk management.

By early 2026, AI infrastructure debt has surged past $200 billion, with projections estimating an additional $800 billion over the next two years, largely financed through complex special purpose vehicles (SPVs) that keep liabilities off hyperscaler balance sheets. For instance, Blue Owl Capital’s $27 billion joint venture with Meta exemplifies how major private credit managers are deeply entangled in this space, raising assets under management to a projected $3 trillion by 2030. This rapid expansion underscores a growing concentration of risk within a few dominant players, amplifying systemic vulnerabilities as the market becomes increasingly interconnected and opaque.

The migration of AI infrastructure debt into mutual funds and retirement accounts, facilitated by giants like BlackRock and PIMCO, introduces significant liquidity and transparency challenges. These funds employ redemption gates and varying valuation methodologies—ranging from daily mark-to-market to quarterly appraisals—that can mask underlying credit deterioration, reminiscent of subprime mortgage dynamics. As redemption pressures mount, with major funds triggering gates throughout 2026, the inability to swiftly liquidate illiquid AI loans heightens systemic risk, potentially catching investors off guard when market repricing accelerates due to delayed returns or operational setbacks.

The reliance on SPV financing structures for AI data centers creates a precarious repayment model dependent on future revenue streams from paying customers, injecting uncertainty akin to the 2008 subprime mortgage crisis. CoreWeave’s approach of isolating loans within shell entities obscures true exposure, while the broader ecosystem’s dependence on timely demand and ROI realization means that any delay—exacerbated by power constraints or permitting issues—could trigger rapid repricing and default cascades. As one analyst warned, 'when data center debt stops being issued, that will be when it's bedtime for this industry,' highlighting the critical role of continuous debt issuance in sustaining the AI infrastructure boom.

The concentration of AI infrastructure debt and insurance within a handful of large funds and off-balance-sheet vehicles not only obscures risk but also mirrors the structural vulnerabilities seen in past financial crises. As Apollo noted, bond coverage ratios for hyperscaler debt have declined sharply—from nearly five times oversubscribed in February to below two times by July—signaling waning investor appetite and potential liquidity crunches. This, combined with the proliferation of covenant-lite loans to weaker credit tenants and the mismatch between long-term data center leases and shorter compute contracts, compounds systemic risk, threatening to unravel the tightly wound financing ecosystem if AI returns falter or are delayed.

By early 2026, the private credit market is witnessing a pronounced pivot toward asset-backed lending, especially within investment-grade segments such as prime and near-prime consumer loans, where supply-demand dynamics and reduced competition contrast with the crowded middle market lending space. This shift is bolstered by technology-driven efficiencies in origination and credit management, enabling managers like Blackstone with its SablePointe Credit Strategies and Ares Management’s $8.5 billion Pathfinder Fund III to capitalize on the growing appetite for asset-based finance (ABF), which Moody’s highlights as offering a broader investable universe and attractive risk-adjusted spreads compared to traditional corporate direct lending.

Amid rising complexity and longer-duration financings, industry leaders emphasize enhanced risk management through rigorous underwriting discipline, collateral quality, and liquidity oversight. Moody’s Marc Pinto underscores that credit performance increasingly hinges on these factors alongside sponsor strength and governance, reflecting a strategic convergence where asset-backed and cash flow lending are blending into hybrid structures featuring both collateral pools and cash flow covenants. Clark Briner further stresses the importance of educating investors on the structural and risk distinctions between asset-backed and cash flow lending, noting that asset-backed finance’s hyper-diversification and short-duration pools provide superior downside protection and liquidity in a market where cash flow lending inflows have surged disproportionately.

Limited partners are recalibrating their focus from pure return chasing to prioritizing downside risk, liquidity, and market correlation, driven by geopolitical uncertainties and credit quality concerns exacerbated by AI-driven disruptions. This risk-conscious mindset fuels growing interest in distressed and special situations credit strategies, which now constitute a significant portion of the private credit universe—nearly 46% outside traditional direct lending. Industry voices like Brad Demong highlight how repricing following earlier 2026 dislocations and software sector stresses have catalyzed demand for these through-the-cycle strategies, as investors seek to capitalize on market dislocations while managing downside exposure.

Technology remains a critical enabler in this evolving landscape, with AI and data analytics powering real-time risk management and fraud detection across credit platforms. Yet, human judgment retains its irreplaceable role, as exemplified by innovations like investment committee AI agents that enhance memo quality and decision-making rigor. This blend of cutting-edge automation and seasoned expertise supports the sector’s strategic shift toward more sophisticated asset-backed and specialty finance lending, ensuring that risk is managed dynamically amid liquidity pressures and underwriting challenges.

AI infrastructure capital expenditures by hyperscalers are skyrocketing, with KB Securities projecting a staggering $725.1 billion spend in 2026—nearly five times the 2023 figure—driven largely by GPU costs that comprise up to 70% of CAPEX. While companies like Microsoft, Alphabet, and Meta maintain several years of funding runway supported by robust 30% AI revenue growth, Oracle stands out with a tighter 1.9-year cushion and a debt/EBITDA ratio of 5.6, signaling mounting credit concerns amid this capital-intensive surge.

The bond markets are feeling the strain as hyperscalers have issued $244 billion in bonds in 2026 alone—more than double 2025’s total—stretching investor appetite and widening credit spreads for giants like Google, Amazon, and Meta. This flood of AI-related debt competes for the same global savings pool needed for critical sectors including federal deficits and infrastructure, raising sustainability questions as hyperscalers edge toward spending more on capex than free cash flow, a shift that challenges their historical reputation as capital fortresses.

Energy costs are a pivotal driver behind rising AI capex budgets, with data centers’ electricity consumption projected to more than double to 945 terawatt-hours by 2030, accounting for nearly half of U.S. electricity demand growth this decade. Coupled with oil prices surging above $100 a barrel, these inflationary pressures exacerbate borrowing costs, exemplified by Meta’s upcoming $12 billion Texas data center financing priced higher than prior deals, and Oracle’s credit default swaps hitting multi-year highs as investors hedge against AI infrastructure debt risks amid geopolitical and rate uncertainties.