AI Salience and the Cost of Equity Capital

This paper examines whether AI-related corporate narratives are associated with firms' implied cost of equity capital (COE). Using U.S. earnings conference call transcripts from 2002 to 2021, we measure AI disclosure with an expanded dictionary generated using the word2vec method and decompose it into a capability-related component (AI capability) and a residual narrative component, which we interpret as AI salience. We develop a stylized framework in which AI capability and AI salience can have different pricing implications: greater capability may increase exposure to AI-related systematic risk, whereas moderate narrative salience may broaden investor recognition and reduce financing costs, although excessively strong salience may trigger skepticism and overstatement concerns. Consistent with this framework, we find that raw AI disclosure is negatively associated with COE. After decomposition, AI capability is positively associated with COE, whereas AI salience is negatively associated with COE. We also find suggestive evidence of non-linearity, with the negative association weakening at very high levels of AI salience. Additional analyses provide complementary evidence consistent with an institutional-investor channel and show that the salience effect is attenuated when market-wide attention to AI rises. Overall, the results suggest that AI-related narratives matter for financing conditions not only because they convey information about underlying technological capability, but also because they shape investor recognition and market pricing.