The grid was not designed for the pace of hyperscale data center deployment, workloads that can double a facility’s power draw with a software update, or demand growth that moves on a timeline measured in months while the infrastructure required to serve it is measured in years.