Preprint2026-11-21

The Investment Cycle Before the Investment: Commitments, Financing, and the Timing of AI Infrastructure

Fuli Yang, Jing Huang, Tingting Qiu, Shiyu Lin

21 November 2026 manuscript

Abstract

Infrastructure investment spans several cohorts. A completion recorded today can reflect an earlier decision to build, and a debt issue can precede access to the proceeds. We track physical milestones in project records and test a spending forecast in issuer accounts. Among 90 Irish data-center-related planning applications with three years of follow-up, one had a linked first compliance certificate within a year and 23 did within three years. Both counts survive a restriction to applications filed after the start of building-control coverage and a search for additional links. A previously documented US campus financing illustrates a third clock: bond closing preceded the instruction to release escrowed funds by 100 days. In a separate issuer-level test, adding growth in seller remaining performance obligations raised pooled squared forecast error by 14.3 percent across 58 two-quarter-ahead forecasts of cash capital expenditure growth. Oracle’s forecast dated September 10, 2025, contributed more excess squared error than the net loss across all 58 forecasts. Together the results establish a distinction essential to dating the investment cycle: later-stage activity can document execution of earlier decisions even as the flow of fresh commitments remains unobserved.

Keywords

Corporate investmentorder backlogproject financedata centersinvestment timing

JEL E22, G31, G32

Cite this version

Fuli Yang, Jing Huang, Tingting Qiu, Shiyu Lin (2026). The Investment Cycle Before the Investment: Commitments, Financing, and the Timing of AI Infrastructure. GARCH Press. https://doi.org/10.69057/WP-0133-V1