Catenaa, Sunday, February 15, 2026-Rapid adoption of artificial intelligence across the United States could add nearly 900,000 tons of carbon dioxide emissions each year, according to a new academic study assessing the technology’s energy demands.
Research published in Environmental Research Letters estimates that broader use of AI systems in business, industry and public services would increase annual US emissions by about 896,000 tons.
The authors said the rise equals roughly 0.02% of total national emissions, indicating a measurable but limited climate impact.
The study examined how AI could be deployed across multiple sectors and calculated the added electricity demand required to support data processing, model training and automated systems.
Researchers estimated energy consumption could rise by up to 12 petajoules annually, an amount comparable to the yearly electricity use of about 300,000 US households.
The findings suggest AI’s carbon footprint remains smaller than that of many energy-intensive industries, even as adoption accelerates.
Still, the researchers said the added emissions highlight the need to improve efficiency as AI becomes embedded across the economy.
The analysis argues that future gains in productivity and automation could be paired with lower emissions if developers prioritize energy-saving hardware, optimized software and cleaner power sources. Without those steps, the cumulative impact of AI-driven electricity demand could grow as systems scale.
The study was authored by Anthony R. Harding and Juan Moreno-Cruz and released in November. The researchers said their work is intended to inform policymakers and industry leaders as AI investment expands nationwide.
Funding support for the research came from Google.
