Hyperscale Data Center Emissions and Energy Use by US State and Region

Supplying aggregated energy consumption and carbon emission metrics for hyperscale data centers across the United States, authored by Gianluca Guidi of Harvard in 2026. It contains 7 files covering state-level totals, balancing-authority fuel mixes, and synthetic facility-level modeling inputs. The data supports research submitted to Nature Sustainability regarding the environmental impact of large-scale computing infrastructure.

Use Cases

• Calculating regional carbon intensity using annual electricity (TWh) and CO2 emissions columns
• Modeling data center capacity workflows using the synthetic all_facilities.csv table
• Visualizing energy grid compositions via the fuel_mix_hyperscalers.csv file

Strengths

• Aggregated at the balancing-authority level for grid-specific analysis
• Includes GeoJSON polygons for EPA regions and balancing authorities
• Derived from peer-reviewed research methodology

Limitations

• Facility-level records are synthetic/anonymized rather than real-world observations
• Lack of exact geographic coordinates for specific data centers due to privacy constraints

Provenance

Source

Harvard T.H. Chan School of Public Health

Collection Method

Aggregation of facility-level data under DUA and generation of synthetic demonstration inputs

Freshness

Last updated March 2026.

Geography

United States