Chasing US financial instability: spillovers between financial conditions and the AI market

Purpose This article examines spillovers between US financial conditions and artificial intelligence (AI)-related equity performance, amid debate over AI-driven financial fragility reminiscent of the 2000 dot-com bubble. Design/methodology/approach The analysis combines the three subcomponents of the Chicago Fed National Financial Conditions Index (credit, leverage and risk) with the S&P Kensho Artificial Intelligence Enablers & Adopters Index, used as a proxy for US AI market performance, over the period from June 1, 2018 to December 12, 2025. It applies the quantile-on-quantile connectedness framework of Gabauer and Stenfors (2024), with a time-varying extension and bootstrap inference. Findings The results show that connectedness between the AI market and financial conditions is state-dependent and becomes intense when the receiving side of the system is in an extreme regime. Credit and leverage conditions affect the AI market asymmetrically and dominate transmission when AI valuations themselves move into tail states. The risk component emerges as the most persistent transmitter across quantile combinations, indicating that changes in volatility, uncertainty and risk appetite are the main drivers of the AI-finance nexus. Feedback spillovers from the AI market appear episodic and arise mainly in cross-tail configurations, when financial conditions are sharply tightening or easing. Research limitations/implications The AI market should be viewed neither as an autonomous speculative sphere detached from macro-financial conditions nor as a passive recipient of broader financial shocks. Rather, it evolves within an asymmetric and regime-dependent system in which financial conditions tend to dominate when AI market returns themselves enter extreme states. By contrast, AI-related valuations feed back into the broader financial environment only under more selective configurations. Practical implications US regulators should monitor periods when high AI valuations coincide with low risk premia, easy leverage conditions and weaker credit quality, using regime-sensitive scenario analysis and stress tests. The objective is to detect not only direct valuation losses, but also indirect effects through lower collateral values, refinancing difficulties, margin pressures and broader balance sheet adjustment. US macro-prudential authorities should also monitor the channels through which AI-related valuations may become systemically important, including market concentration in a small number of AI-exposed firms, the role of high valuations in supporting collateral and borrowing capacity, and the degree of refinancing dependence. Originality/value This article is one of the first attempts to study US AI market evolution and financial conditions from a macro-financial outlook. It contributes to the emerging AI-finance literature by introducing a financial-conditions perspective into the analysis of AI-related valuations. It also extends the literature on AI-related risk transmission by showing that the relationship between the AI market and macro-financial conditions is nonlinear, asymmetric and regime-dependent. Methodologically, it is among the first applications of the quantile-on-quantile framework, extending Gabauer and Stenfors (2024) by introducing a time-varying dimension and bootstrap inference.