What happened
A new BIS Working Paper models the current AI capital expenditure boom as a 'dynamic contest' among a handful of firms racing for dominant market positions, and finds the race generates severe over-investment relative to the socially efficient level. The headline quantitative finding: 'the model points to over-investment of around 1.5 times the efficient level, rising to around three times where demand is less elastic.' The paper, calibrated to company balance-sheet and disclosed-deal data, argues that debt financing and circular equity stakes between AI firms introduce financial fragility, exposing the sector to fire sales of specialised (non-fungible) AI hardware assets if revenue disappoints. A network analysis further shows that the failure of a single major AI firm could cascade through chains of financial exposure to other firms, especially in a concentrated network — a systemic-risk framing rarely applied to the AI buildout by a central-bank body.
Why it matters
This is one of the first quantitative, central-bank-grade financial-stability analyses of the AI capex boom, giving boards and CFOs a concrete over-investment multiplier and a contagion mechanism to weigh against continued AI infrastructure commitments and financing structures (debt, circular deals).
Action needed
Brief the board and finance committee on AI capex financing exposure and circular-deal concentration risk; stress-test balance-sheet exposure to a potential AI infrastructure repricing.