Harvard study: just 403 US data centers operational between May 2024 and April 2025 consumed as much electricity annually as the state of Michigan
First reported by Wired ·
The electricity required for data centers, especially those powering AI, now rivals entire states, directly increasing the proportion of energy drawn from fossil fuels and raising carbon emissions significantly.
A Harvard study published in Nature Sustainability estimates that 403 large US data centers operational between May 2024 and April 2025 consumed as much electricity annually as the entire state of Michigan. The study revealed that the majority of this energy was sourced from fossil fuels, resulting in data center emissions tripling since the start of the AI boom. Current federal and state tracking methods do not distinguish data center energy consumption from the broader power sector, and specific usage figures are considered proprietary. The research focused on hyperscale facilities, which are the largest data centers and have faced community opposition, although not all surveyed were solely used for AI. Researchers compiled their data using industry research and satellite imagery, modeling energy use and regional grid mixes to estimate carbon emissions. Facilities in Virginia, Oregon, Ohio, and Iowa accounted for over half of the study's energy consumption.
The study highlights a critical information gap: the precise energy and emissions footprint of data centers remains largely obscured due to proprietary data and a lack of dedicated government tracking. This opacity makes it challenging to accurately assess the strain on power grids and to enforce environmental regulations, especially as the demand for computing power, driven by AI and other applications, continues to surge. The reliance on fossil fuels for a majority of this energy consumption points to a growing environmental concern that current infrastructure and energy policies may not be equipped to handle.
The rapid growth of hyperscale data centers, even predating the current AI boom, signifies a fundamental increase in demand for electricity that is outpacing our understanding and management capabilities. The emergence of behind-the-meter power generation, often utilizing inefficient natural gas turbines, further complicates emission assessments and suggests a potential for significant, unmonitored increases in overall US emissions. As the Energy Information Agency pilots new voluntary surveys, the industry's future energy consumption and environmental impact will remain a key area to watch, with implications for grid stability and climate targets.
AI-written summary. May contain errors.