The Sequence Opinion - Issue #961: The AI Credit Crunch
How the financing supply chain works and why lenders are getting nervous
TL;DR
- AI companies face a gap between upfront infrastructure costs (GPUs, buildings, electricity) and revenue generation, which is bridged by credit.
- Lenders are becoming more selective, demanding higher returns or more protection, as evidenced by a CoreWeave bond yielding around 15%.
- Factors contributing to financing difficulties include construction and power delays, aging hardware, loans outlasting customer contracts, concentrated customer bases, and increased cost of capital.
- Meta's joint-venture and equity funding approach and SpaceX's xAI GPU lease commitment are presented as case studies.
- The bear case involves disappointing cash flows and cascading damage through suppliers, customers, and financiers.
- The bull case relies on strong sponsors, equity buffers, and the continued utility of older hardware.
- Credit availability is expected to be a deciding factor in AI infrastructure development, with key questions focusing on cash collection, principal repayment, contract renewals, and responsibility for construction delays.