Can AI computing power promote the green transformation of energy enterprises? Evidence from the nonlinear moderating effect of public environmental awareness

Amid growing concerns over climate change and energy security, the green transformation of energy enterprises has become a global sustainability priority. This study investigates the nonlinear impact of artificial intelligence computing power (AICP) on the green transformation of 251 Chinese energy enterprises from 2010 to 2023. Results reveal a statistically significant U-shaped relationship: AICP initially inhibits green transformation due to high costs and inefficiencies but later promotes it by enhancing environmental and operational performance. Public environmental awareness is found to moderate the nonlinear relationship. A high levels of concern can alter the curvature of the U-shape, sometimes leading to symbolic rather than substantive green actions, thereby weakening AICP's positive effects. Heterogeneity analysis shows the U-shaped effect is stronger in non-state-owned enterprises and in firms located in low-carbon pilot cities, indicating that ownership structure and regional policy context shape AI-driven sustainability outcomes. Additionally, threshold regression results reveal that when the carbon disclosure index exceeds a critical value, the relationship between AICP and green transformation becomes an inverted U-shape. These findings reveal that the interplay between technological innovation, public expectations, and institutional environment is critical for designing targeted policies that unlock the full green potential of AI in the energy sector. • This study identifies a significant U-shaped relationship between AI computing power and the green transformation of energy enterprises, revealing both the risks and potentials of digital technology in sustainability transitions. • Public environmental awareness plays a nonlinear moderating role , which may attenuate the effectiveness of AI in promoting green transformation due to symbolic responses and legitimacy pressures. • A threshold regression uncovers a turning point beyond which increased carbon disclosure leads to a shift from a U-shaped to an inverted U-shaped relationship, highlighting the importance of institutional transparency in shaping AI's environmental returns.