The $1.65 Trillion AI Debt Trap

This is a funding-cost story more than a pure AI demand story. Once growth depends on SPVs and private credit, the market stops valuing these names like software and starts valuing them like levered infrastructure: small changes in credit spreads can drive outsized multiple compression, even before reported earnings move. ORCL is the most fragile because its equity is more exposed to refinancing optics and to any forced repricing of future AI capacity.

The second-order loser is the AI supply chain. If hyperscalers become more selective on financed deployments, the first hit shows up in order books for NVDA-linked compute, networking, and power infrastructure before it shows up in revenue. By contrast, MSFT and GOOGL are better positioned to absorb higher funding costs internally and can use balance-sheet strength to take share from weaker cloud/AI peers, especially if lenders demand tighter covenants or more equity support.

Near term, this is mainly a volatility and multiple event; over 1-3 months, watch for CDS/bond follow-through and any language shift around capex discipline. Over 6-18 months, if AI buildout continues to require private credit, terminal multiples for the whole basket should compress as investors price in slower payback and higher leverage. The key falsifier is a stabilization in AI credit spreads plus unchanged or higher capex guidance, which would imply the market is overpricing the funding risk.