Conference Agenda

Artificial intelligence (AI) interacts with climate in various ways. Its impact on emissions and the costs and benefits of emission reduction are uncertain. A critical challenge is to align AI investment with global climate policies. We propose a unified framework, integrating AI's impact on emissions, economic production, and climate damages. We extend the DICE model to a dynamic general equilibrium climate-economy model with endogenous AI growth. By boosting both labor productivity and capital deepening, AI investment increases future gross economic output, and so emissions. It also requires more energy- and emission-intensive capital. At the same time, the application of AI technologies lowers the carbon intensity of the broader economy. Moreover, AI-powered adaptation alleviates climate damages. We distinguish between ICT-like and Industrial Revolution (IR)-like prospects for AI development. Calibrated to the best available evidence, we find that AI development is net polluting. Under current low abatement, ICT-like AI raises global warming further by an additional 0.1°C in 2100, while IR-like AI raises it by 0.8°C. The associated climate costs are 1.6% and 23% in permanent consumption loss, offsetting roughly one‑fifth and one‑quarter of AI’s economic gains, respectively. Consequently, the optimal AI investment rates are reduced from its maximum potential---for ICT-like AI, by 0.12 p.p. to 1.96% in 2050 and by 0.47 p.p. to 3.25% in 2100. Notably, proactive climate policy can largely reverse this investment decline, indicating that it is a complement to, rather than a substitute for, AI development. We further show that the trade-off between AI and abatement investment is driven primarily by their relative marginal economic returns, not by AI's emission impact.