AI investment race leads to over-commitment, financial fragility

The AI build-out represents one of the largest technology-driven investment booms in US history, with total capital expenditure by major hyperscalers set to exceed $700 billion in 2026 alone.

The paper models a dynamic contest where firms competing for a few dominant positions over-commit resources, leading to over-investment.

This over-investment is estimated at around 1.5 times the efficient level, rising to approximately three times where demand for AI services is less elastic.

The larger the boom, the deeper the eventual bust, as the sector becomes exposed to revenue disappointment.

This tendency toward excessive commitment raises questions about the sustainability of the current boom, which is on track to outgrow every previous episode only three years in.

The financing of the AI boom displays vulnerabilities seen in earlier boom-bust episodes, with some firms increasingly resorting to debt through bond issuance, special purpose vehicles, and non-bank lenders.

Circular financing, where a firm takes an equity position in an AI lab in exchange for compute commitments, creates interconnectedness and exacerbates losses during busts.

The paper's central result is that the AI boom creates endogenous fragility: the more capacity built, the higher the productivity bar to sustain it.

A network analysis reveals that stress in one firm could cascade to others through chains of financial exposures, especially as funding concentrates on backers bridging several labs.