Name
Open-source Generative AI and Firm Value: Evidence from the Release of DeepSeek
Date & Time
Sunday, July 5, 2026
Description

In this paper, I exploit the release of DeepSeek in January 2025 as a shock to the availability of open-source Generative AI (GenAI) models and investigate the value implications for corporate GenAI adopters. I document that U.S. firms with high pre-event GenAI exposure, i.e., those that have adopted or plan to adopt GenAI, earn an average cumulative abnormal return of 1.2% over the event window, relative to low-exposure firms. The effect is stronger for firms with tighter financial constraints and for those with proprietary information concerns, consistent with the benefits of open-source GenAI in reducing adoption costs and mitigating privacy concerns. In the post-event period, high-exposure firms are more likely to articulate adoption plans and benefits of DeepSeek during conference calls, and to embed DeepSeek in their algorithms shared on GitHub. High-exposure firms also experience an upward revision of analyst forecasts and a more positive media tone. In contrast, GenAI providers and their hardware suppliers experience negative abnormal returns. Lastly, the baseline analysis for Chinese firms yields a larger market impact of DeepSeek’s release. Overall, the findings indicate that open-source GenAI can further unlock the valuation potential of adopting GenAI.

Yue Zhao
Keywords
Generative AI, DeepSeek, Open-source, Proprietary Information, Firm Value
Theme
FINANCIAL ACCOUNTING
Author 1
Yue Zhao