Firm-level evidence on AI-driven output expansion and productivity in China

Artificial intelligence (AI) has the potential to transform productivity across industries, yet firm-level empirical evidence remains limited in emerging economies. This paper examines the impact of AI adoption on firm-level output growth and total factor productivity, using data from Chinese listed companies between 2009 and 2021. We measure AI investment through firm-level spending on software, cloud services, intellectual property, and advanced digital technologies disclosed in financial statements. This measurement approach reflects AI’s role as a general-purpose technology increasingly embedded in digital infrastructure and business processes—features that are not well captured by traditional proxies such as industrial robot usage. To address endogeneity, we employ a nonparametric production function estimation with instrumental variables, using regional variation in digital economy policy as an exogenous source. Results show that AI investment significantly boosts output, particularly in light manufacturing, chemicals, and high-tech sectors. While intermediate materials remain the primary input, AI’s contribution to aggregate output growth has steadily increased. AI adoption also enhances firms’ resilience during downturns, though the benefits are uneven—large firms gain substantially, whereas small and medium enterprises see more modest effects. Further analysis reveals short-run implementation costs that can temporarily reduce productivity, though persistent AI adoption yields divergent long-run outcomes across industries. Quantile regressions show that lower-productivity firms often realize initial gains that fade or reverse, while frontier firms enjoy sustained improvements. Finally, we identify strong positive spillovers: AI investments by nearby firms generate external productivity gains, highlighting the importance of innovation clusters. Overall, our findings position AI as a key driver of output and productivity in emerging economies, and emphasize the need for targeted, inclusive policy frameworks to support its widespread and equitable adoption.