Financing AI without a SaaS repeat

AI compute financing is likely to create a larger, more speculative and more correlated credit cycle than SaaS, requiring concentration limits before the buildout accelerates.

• Software attracted direct lenders because it combined scalable equity upside with cash flows that appeared sticky and uncorrelated.
• Generative AI changed that profile by making technological obsolescence a risk shared across the software sector.
• Bank analysts estimate that private credit could finance roughly $1 trillion of the AI compute buildout, more than half of the industry's current assets under management.
• The financing will use largely untested structures, including construction and power finance, GPU-collateralized asset-backed financing, off-balance-sheet SPV lending, senior secured facilities for neoclouds, and M&A lending.
• AI credit is more speculative because the industry has not settled how spending will divide among hardware, infrastructure, models, and applications, while data-center default rates may move with the broader economy.

Takeaway: Investors should not avoid AI compute exposure outright, but should establish concentration limits in advance because the risks are more correlated and less proven than those of SaaS.