The paper is strongest when mapping the operational friction between fast computational workloads and slow grid infrastructure. Large-scale graphics processing unit clusters create power fluctuations of hundreds of megawatts within seconds. These rapid load swings strain distribution networks and threaten real-time stability.

Yet the paper stops at the data center fence line. For executives running steel mills, hydrogen electrolyzers, and chemical plants, this load growth dictates a harsher reality. AI data center expansion bids up wholesale electricity prices and consumes regional grid capacity. Fifteen states account for 80% of United States data center demand, creating localized bottlenecks. Rising household electricity bills driven by this load growth trigger regulatory pushback, as seen where Regional Greenhouse Gas Initiative auction proceeds help offset consumer price spikes (CleanTechnica, 10 Sep 2026). When data centers capture available transmission and generation assets, heavy industry loses access to the cheap power required to decarbonise steel, hydrogen, and chemicals.

Grid capacity dictates capital allocation timelines. Reliance on traditional utility interconnects introduces delays that break project economics for heavy industry. Future decarbonisation requires dedicated co-located generation to bypass the grid constraints that AI workloads create.

Reference: Chen et al. (2026). Electricity demand and grid impacts of AI data centers: Challenges and prospects. Nexus — https://doi.org/10.1016/j.ynexs.2026.100162