Buying Capacity Before It Exists: Advance Payments and Supply Commitments in AI Infrastructure
Advance payments and supply commitments organize infrastructure investment before projects are ready to operate. Buyers secure future equipment, accept obligations, and provide suppliers with funding while other development tasks remain open. This paper examines that process in the equipment market serving the AI and electricity investment cycle, combining disclosures from GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries with buyer agreements, an unsuccessful project, and a completed historical project. Contract evidence shows that access, price, purchase obligations, and exit rights become effective at different stages. Customer funding, equipment delivery, and commercial operation follow distinct schedules. A stock-flow reconciliation shows how the advancement of these commitments shapes the investment cycle's reported momentum. GE Vernova's quarterly gross additions to firm backlog doubled from 8 to 16 GW between the first and third quarters of 2026, while new commitments entering its combined reservation-and-order pipeline fell from 21 to 11 GW. Conversions supplied 80.6% of gross firm-backlog additions over the nine months. Earlier procurement decisions therefore sustained formal-order growth as new intake weakened. Advance contracting connects the acquisition of future supply to the accumulation of financial and commercial obligations. Reading those obligations alongside new entry, conversion, shipment, and operation explains how an infrastructure boom continues to advance while its sources of momentum change.
Authors
- Fuli Yang (ORCID: https://orcid.org/0009-0005-4055-7561)
- Jing Huang (ORCID: https://orcid.org/0009-0004-8120-1652)
- Tingting Qiu (ORCID: https://orcid.org/0009-0009-2442-5621)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-05
- DOI
- https://doi.org/10.5281/zenodo.23959266
- Primary Topic
- Capital Investment and Risk Analysis
- Type
- preprint