Concentrated siting of AI data centers drives regional power-system stress under rising global compute demand

A team of researchers led by Danbo Chen has published a pivotal study forecasting the massive and concentrated electricity footprint of AI data centers. Using a novel AI-energy coupling framework that combines LLM-based analysis of corporate and policy data with quantitative energy-system modeling, the study projects that aggregate electricity consumption by the six leading AI firms will more than double, from roughly 118 TWh in 2024 to between 239 TWh and 295 TWh by 2030. This staggering growth would make the AI industry alone responsible for about 1% of total global power demand within six years, transforming it from a marginal digital service into a structural component of global energy dynamics.

The study's critical finding is the geographic concentration of this new infrastructure, with over 90% of projected compute capacity located in North America, Western Europe, and the Asia-Pacific. This creates severe regional grid stress. The researchers developed a Power Stress Index (PSI), identifying specific regions like Oregon, Virginia, and Ireland as highly vulnerable, with PSI values exceeding 0.25. In contrast, more diversified and resilient grid systems, such as those in Texas and Japan, are better equipped to absorb the new load. The paper concludes that the current trajectory poses significant risks to grid stability and underscores an urgent, global need for anticipatory infrastructure planning that directly couples computational growth with accelerated renewable energy expansion and grid modernization to ensure resilience.