Expected Economic Impacts of Open-Source AI Models: Firm-Level Evidence

(co-authored with Yi Huang, Sung Kwan Lee, Yujun Zhou)

Abstract

We examine firm-level stock-market reactions to AI model releases and find that opensource releases generate more favorable valuation responses than proprietary releases. Relative to proprietary releases, open-source releases disproportionately benefit smaller, capital-constrained firms, consistent with their greater accessibility and lower user costs. Valuation gains after open-source releases concentrate among firms with greater exposure to AI-suitable tasks, particularly those involving human–AI augmentation; comparable patterns are absent after proprietary releases. Using input-output exposure measures, we further document positive valuation spillovers from high-AI-exposure firms to upstream and downstream partners in low-AI-exposure industries, but only following open-source releases. Taken together, our findings suggest that open-source AI broadens the expected economic benefits of technological progress by lowering adoption barriers and extending benefits across firms and interconnected industries.