Rapid expansion of Artificial Intelligence is expected to drive proportional expansion of economic activity through productivity gains–potentially leading to higher energy consumption and associated environmental impacts, including increased carbon dioxide emissions. We combine data on economic activity with early estimates of likely adoption of AI across occupations and industries to quantify the potential change in energy use and carbon dioxide emissions for the United States. At the industry level, we estimate annual increases in energy use ranging from 0 and 12 petajoules (PJ) and carbon emissions from 0 tonnes to 272 kt (ktCO2). Aggregated across the economy, AI adoption could lead to an additional 28 PJ of energy use and 896 ktCO2 in emissions annually–equivalent to approximately 0.03% of annual national energy use and 0.02% of annual national CO2 emissions. These results highlight the need to account not only for the aggregate energy and environmental implications of AI-driven productivity gains, but also for how these impacts vary across industries based on their specific characteristics.