A new study published in the journal Nature Sustainability reveals that hyperscale data centres in the United States consume more electricity than many midsized nations. The research indicates that fossil fuels remain the primary power source for these massive facilities, which are largely being constructed to support generative artificial intelligence. Consequently, the sector contributes a significant volume of greenhouse gas emissions to the global climate.
This finding highlights a critical disconnect between the energy demands of modern computing and the transition to renewable power sources. While industry rhetoric often focuses on the efficiency gains of new models, the physical infrastructure powering those calculations relies heavily on existing coal and gas grids. The study suggests that reducing the carbon footprint of artificial intelligence requires more than just software optimisation; it demands substantial investment in grid decarbonisation and power generation. Without addressing the source of this electricity, the environmental costs of scaling these systems will continue to rise.
- Fossil fuels still supply the majority of power for US hyperscale data centres.
- Total energy consumption by these facilities exceeds that of several countries.
- Significant work remains to reduce pollution linked to AI infrastructure.




