The only thing keeping the big AI labs from going broke right now is basically "faith" that they’ll eventually build a super-intelligence.
SSRN · March 13, 2026 · 6305300
Why it matters
This economic model reveals that the pursuit of AGI creates a 'faith-based survival' effect: because future capabilities are expected to be so valuable, they lower a firm's current default risk. This allows labs to maintain massive overinvestment and debt that would be considered irrational in any other industry.
From the abstract
Frontier AI laboratories racing toward Artificial General Intelligence (AGI) face uncertain timing of transformative demand and must decide how much irreversible capacity to build and how to split it between inference (current revenue) and training (future capability). I develop a real-options model with regime switching, duopoly competition, endogenous default risk, and diminishing returns calibrated to AI scaling laws. The model delivers analytical investment triggers and a duopoly preemption