The escalating energy consumption by AI-driven datacenters has raised concerns, as projections suggest demands will outstrip chipmakers’ ability to keep pace. Deloitte Insights predicts US datacenter power needs might quintuple by 2035, prompting Bain & Company to advise utility firms on scaling energy resources. However, an overestimation would burden customers if infrastructure expansions go unused.

The Southern Environmental Law Center enlisted London Economics International to scrutinize this issue. Their report reveals that datacenter builders often duplicate grid connection requests, leading to inflated demand forecasts. Moreover, the anticipated AI expansion may not materialize as chip supplies are already stretched thin.

Notably, the US would require 90% of global AI chip growth to meet its datacenter goals, a scenario deemed unrealistic. This has implications for energy firms, as over-forecasting could inflate costs for consumers. Some US utilities are already increasing rates due to these developments, with National Grid and Northern Indiana Public Service Company announcing hikes.

Ultimately, the report concludes that accurately predicting the datacenter’s future energy needs is fraught with uncertainty. The risk of overestimating is significant, risking financial repercussions for existing energy consumers. Reuters also notes that PJM Interconnection, serving eastern states including Virginia, is poised to raise energy rates by over 20%.