Simulation finds more sites for flexible data centers

A data center drawing 1 or 2 gigawatts would put an unusually heavy load on a power grid. In a Texas-scale simulation, Harvard electrical engineer Le Xie and colleagues found that more sites could host facilities of that size when operators agreed to cut or shift electricity use during periods of grid stress.

The approach changes the order of operations. Instead of letting developers choose locations and then wait in an interconnection queue while utilities assess the grid, planners would identify suitable sites in advance and set operating limits for large data centers. The model found that this flexibility increased the number of viable locations by about 9% to 21% without meaningfully raising average power prices.

The pressure is growing alongside AI computing. The researchers tested loads much larger than those of a typical data center, but in line with some of the biggest campuses now being planned. A 1-gigawatt facility can draw roughly as much electricity as 800,000 average American homes; a 2-gigawatt facility can draw twice that amount.

And then what, concretely? If the simulated approach holds up beyond the model, grid planners could make room for some large data centers without requiring major transmission upgrades that could compete with the broader needs of the energy transition. The authors say connection times could potentially fall from roughly five to eight years to 12 to 18 months.

The evidence still has a firm boundary. The results were produced on a synthetic grid designed to resemble Texas, not on the real Texas power system, so the paper is a proof of concept. Former Energy Secretary Jennifer Granholm called the shift toward considering flexibility before projects enter the queue a “pretty profound shift,” but the model's findings still need testing in real-world planning and operations.