How Do Utilities Determine Which AI Data Centers Get Grid Access?

Utilities must decide which artificial intelligence data center projects are substantial enough to justify billions of dollars in grid infrastructure investments. Rather than assuming all announced projects will materialize, utilities are using different strategies to filter demand: some apply risk-adjusted forecasting that runs simulations to account for delays and cancellations, others model multiple hypothetical scenarios to identify transmission needs across different futures, and still others categorize projects by contractual commitment levels. The challenge is significant because overestimating demand risks building infrastructure for projects that never happen, while underestimating it could leave data centers waiting years for necessary upgrades. These planning decisions reflect a broader shift where utilities are expanding their traditional forecasting tools to handle the unprecedented pace, scale, and uncertainty of large power requests from data centers.