Debt Has Entered the A.I. Boom (www.nytimes.com)

Blackstone is preparing a $3.46 billion commercial-mortgage-backed securities (CMBS) deal to refinance QTS Data Centers’ assets — 10 hyperscale facilities across six U.S. markets — marking the largest data-center CMBS offering this year. The bonds would be collateralized by sites that consume enormous power (enough to run Burlington, Vt., for five years) and come amid a surge in capital spending: McKinsey estimates $7 trillion in data‑center investment needed by 2030, and hyperscalers logged roughly $112 billion in capex in the past three months. More broadly, data‑center asset-backed securities (ABS) issuance has jumped (about $13.3 billion across 27 deals this year, +55% vs. 2024), as cloud giants diversify financing beyond internal cashflow. Technically, the market is shifting toward complex structures—S.A.S.B. (single-asset, single-borrower) deals concentrate risk in one operator or facility, and special-purpose vehicles (SPVs) let companies push debt off their balance sheets (used by Meta and emulated by others like xAI). That raises two key implications for AI/ML infrastructure: it unlocks massive funding to scale compute quickly, but concentrates credit and obsolescence risk (e.g., stranded assets if chip architectures change or revenue fails to materialize). Regulators and banks (including the Bank of England) warn the transition from cash-funded builds to debt-heavy financing could amplify systemic risk, making investor due diligence and technology‑resilience clauses increasingly important for the AI ecosystem.