AI demand strains power grids and data center capacity

The rapid expansion of artificial intelligence is straining the physical infrastructure required to support it, creating a critical bottleneck defined by shortages of power, space, and specialized networking equipment. As AI models grow larger and more complex, the demand for electricity and data center capacity is outpacing the ability of utilities and developers to supply it, forcing a fundamental shift in how companies and investors approach the sector.

A new report from the Capgemini Research Institute, based on a global survey of over 600 senior electricity executives conducted in January 2026, underscores the scale of the challenge. The study found that 77% of utilities are struggling to accurately forecast future electricity demand due to the volatile and unpredictable nature of AI workloads. This uncertainty is exacerbated by what executives describe as "phantom" data center load requests, with approximately 19% of projected demand never materializing, distorting planning and increasing the risk of both overinvestment and underinvestment.

The core issue is that AI training and inferencing consume vast amounts of power, with consumption expected to rise to between 25% and...