Working Paper 10
AI Capex and the Corporate Credit Market: Evidence on Crowding Out, 2010-2026
September 2026
Abstract
We ask whether the rapid increase in AI, datacenter, and hyperscaler financing since 2023 is changing the allocation, price, and availability of credit for other corporate borrowers. Measured AI-side investment, defined as cash capital expenditure of the four hyperscalers plus Oracle and CoreWeave, was $389 billion in 2025, up from $24 billion in 2015, with about $789 billion guided for 2026. Against that boom, US corporate credit aggregates expanded rather than contracted: bond issuance averaged $1.90 trillion per year in 2023-2025 versus $1.68 trillion in 2015-2022, bank credit reached $19.8 trillion in August 2026, and credit spreads narrowed. In the euro area, total bank lending to non-financial corporations grew 1.7 percent per year in 2023-2026 versus 2.0 percent in 2015-2022, and the reconstructed credit premium was flat at about 1.9 percentage points. What the data do show is sectoral divergence rather than aggregate rationing: euro-area manufacturing loan stocks contracted after 2022 while real estate and professional services kept growing, technology borrowers absorbed a rising share of growing US credit, and AI financing increasingly flows through private credit, project vehicles, and foreign-currency bond markets that sit partly outside measured bank aggregates. Every result is descriptive; the samples are short and the post-2023 window coincides with the monetary tightening cycle, so nothing here identifies a causal effect of AI investment on credit conditions.