The study, hosted by Harvard researchers, evaluates the energy consumption and carbon output tied to the nation's largest cloud facilities. According to the analysis, these massive campuses consumed 93.66 terawatt-hours (TWh) of electricity, with nearly 60% of the resulting emissions originating from fossil fuel generation. The 52.69 million metric tons account for approximately 1.10% of total U.S. carbon emissions linked to electricity consumption in 2023.
Study authors emphasize that the calculated figure represents a conservative lower bound. The 403 analyzed facilities account for roughly 78% of estimated U.S. hyperscale data center electricity use for 2024, leaving out smaller edge sites, enterprise server rooms, and the remaining slice of hyperscale infrastructure. Furthermore, the total is estimated to be more than five times higher than comparable emissions estimates from 2018, reflecting rapid expansion across the sector.
While the Denmark comparison provides scale shorthand for the industry's total footprint, analysts note that the exact statistical basis for the country comparison remains broad, serving primarily to illustrate the sheer volume of power drawn by modern cloud and AI infrastructure.

