A Multi‐Stage Efficiency Analysis of Taiwan's Semiconductor Industry: Integrating ESG and Artificial Intelligence Perspectives

ABSTRACT The semiconductor industry is central to global technological progress but increasingly faces the challenge of balancing financial performance with rising environmental, social, and governance (ESG) expectations. While ESG and artificial intelligence (AI) are widely regarded as strategic priorities, limited empirical research has investigated their combined influence on firm‐level efficiency in high‐tech, capital‐intensive sectors. This study addresses this gap by applying a three‐stage Data Envelopment Analysis (DEA) to evaluate the performance of 33 publicly listed semiconductor firms in Taiwan across profitability, sustainability, and market value. The model reflects the sequential value‐creation process: the Profit stage measures how R&D and labor are transformed into innovation output (intangible assets and income); the Sustainability stage assesses how governance and environmental inputs shape ESG scores; and the Market stage examines how ESG performance and capital investment affect market valuation. Stage‐specific efficiency is measured using ESG and structural variables aligned with the semiconductor sector's vertically disintegrated and interdependent value chain. AI‐driven innovation is proxied by patent activity. Results reveal performance disparities across stages: Market stage efficiency ranks highest (0.9664), followed by Profit stage, while Sustainability stage lags—highlighting an ESG implementation gap in otherwise competitive firms. Tobit regression analysis shows that AI‐driven innovation enhances sustainability efficiency, whereas firm size and leverage are more closely linked to profitability. This study proposes a multi‐stage framework integrating ESG and AI to evaluate sustainable innovation in high‐tech sectors, offering insights into how firms can balance competitiveness with long‐term sustainability.