Market Reactions to Generative AI Announcements: An Event Study Analysis of Abnormal Returns, 2022–2025

Abstract

Kabir Bhushan*

This paper asks a straightforward question: when a major generative AI announcement drops, do markets actually price it right away, or does something more complicated happen? Using standard event study methodology applied to six AI-linked equities (Nvidia, Microsoft, Alphabet, Meta, AMD, and Palantir) across 18 significant generative AI announcements from November 2022 through January 2025, I find that the answer is mostly the latter. Markets react fast and in the right direction, with announcement-day abnormal returns averaging +1.84% and are highly significant, but they do not get it fully right on day one. A statistically significant positive drift on day +1 (+0.71%, p < 0.05) and persistent post-announcement drift following the largest positive-surprise events suggest that generative AI information is incorporated gradually, consistent with conservatism bias among investors. On the negative side, the Deep Seek R1 release in January 2025 triggered a sector-wide selloff with Nvidia dropping roughly 17% in a single day and the six-firm portfolio averaging a -6.8% five-day CAR, a reaction roughly 24% larger in absolute terms than any positive- surprise event, which is exactly what loss aversion theory would predict. These findings sit uncomfortably with the semi- strong form of the efficient market hypothesis and point to real behavioral mechanisms at work in how investors process transformative technology news. Policy implications for AI disclosure standards and practical takeaways for investors in AI-linked equities are discussed.

HTML PDF