In the U.S., the mismatch between AI-fueled energy demand and supply is clear. Hyperscalers/data center developers especially are more skeptical of the current pace of energy deployment when juxtaposed with future demand.
- Fifty-seven percent (57%) of all respondents agree or somewhat agree that the current pace of energy deployment in the U.S. is not enough to meet energy demand caused by AI.
- Hyperscalers/data center developers are more skeptical of the current pace of energy deployment; 64% of hyperscalers/data center developers agree or somewhat agree that the current pace of energy deployment is not enough to meet energy demand caused by AI, compared to just 44% of electric producers/utilities.
- Seventy-six (76%) percent of all respondents see electricity usage in data centers increasing by at least 10% annually, with 30% indicating they think it will increase by at least 15%. Hyperscalers/data center developers are more likely to predict faster demand growth, with 43% predicting 15% or more.
- The top two challenges cited by electric producers/utilities to connecting data centers to the grid include lack of grid capacity (76%) and unreliable load projections (71%).
“Extreme weather events and the need for more resiliency; growing demand on the grid thanks to a boon in domestic manufacturing, increasing popularity of EVs, and transitions to the cloud – before all eyes were on AI, power and utility companies were already between a rock and a hard place for where to put their resources and capital first,” said KPMG U.S. Energy Leader Angie Gildea.