AI Data Centers Could Integrate with Power Grids Through Flexible Demand, Study Suggests
A recent modeling study by an international research team, led by TU Wien, suggests that AI data centers in the U.S. could operate within existing grid capacities by reducing electricity consumption for as little as 35 hours per year. This approach challenges the traditional view of data centers as constant consumers and proposes they become flexible entities that adapt to grid conditions. The study highlights that many computing tasks, such as training large AI models, can be shifted to times when renewable energy is abundant. This flexibility would require special contracts guaranteeing a base capacity while allowing temporary reductions during grid strain. The research also points out the need for better data on data center electricity consumption to effectively plan grid infrastructure. Currently, the energy demand of AI data centers is dramatically increasing, with 39% of gas-fired power generation capacity under development in the U.S. by the end of 2025 intended to supply these facilities, raising c...
