UBS warns AI-driven shock could lift Swiss private-credit defaults to 13% in a worst-case scenario

UBS constructed stress scenarios that treat accelerated AI deployment — driven by large, GPU‑heavy data‑center projects — as a catalyst for concentrated earnings shocks, swift risk repricing and liquidity squeezes in private‑credit portfolios. In its most severe path, cumulative defaults rise to about 13%, well above normal expectations for these leveraged, illiquid assets. The bank links the risk to a narrow set of borrowers: software vendors with heavy capex exposure to hyperscaler customers, small hardware suppliers reliant on shifting wafer and DRAM allocations, and privately held firms with weak covenant protection. Market evidence complements UBS’s modelling: participants estimate roughly $3 trillion of AI‑focused data‑center projects under consideration, with hyperscaler procurement possibly totaling about $1.5 trillion by 2025, concentrating demand in a few anchor tenants. New financing approaches — from CMBS‑style securitisations and bespoke structured credit to parcelled private placements aimed at technical and energy risk — are emerging to fund the buildout, but they also embed correlated execution and utilization risks. Local permitting delays and community opposition have already put roughly $64 billion of U.S. data‑center projects at risk of delay or cancellation, creating real timing and cash‑flow uncertainty for sponsors and suppliers. UBS highlights transmission channels: markdowns by private funds can trigger redemptions and forced sales into thin secondaries, widening realized losses; public credit desks have begun penalizing software and narrowly focused vendors, raising refinancing costs; and upstream hardware concentration can create delivery timing shocks that materially alter borrower cash flows within one investment cycle. Private‑equity and infrastructure managers are responding by shortening effective holding horizons, tightening covenants, formalizing operational playbooks and modelling accelerated obsolescence and replacement‑cost dynamics — actions that both mitigate and crystallize repricing across similar borrowers. For allocators and lenders, UBS’s work recommends revising liquidity buffers, covenant design and concentration limits and testing manager–LP alignment under faster, sectoral repricing paths. Regulators and bank risk teams should treat the scenarios as a prompt to stress‑test liquidity‑management protocols, redemption terms and sectoral leverage monitoring. UBS does not assert the worst case will occur, but by mapping a plausible chain from AI‑driven capex concentration to private‑credit losses it reframes the debate: AI can be an accelerant of timing and severity in credit events, and market participants need to price and prepare for wider downside and compressed timing risk than traditional credit‑cycle models assume.