ABSTRACT:

This paper examines the macroeconomic and financial effects of artificial intelligence (AI) information shocks, defined as unanticipated revisions in expectations about the economic implications of AI. We identify these shocks in a structural vector autoregression using an external instrument based on abnormal equity returns of AI-related firms around selected events. This instrument measures information intensity through the size of market repricing and includes announcements with both favourable and adverse valuation effects. We find that revisions in expectations about AI affect equity valuations and industrial production differently in the US and the UK, while employment declines in both countries. Beyond equity valuations, financial effects are also reflected in credit markets, as corporate credit spreads narrow and the excess bond premium declines. These adjustments extend to the banking sector, where responses vary according to relative AI exposure. Overall, the results show that revisions in AI expectations shape economic activity and financial conditions.