arXiv:2611.24506v1 Announce Type: new Abstract: Grid connection constraints increasingly limit the expansion of AI data centers, yet the additional hosting capacity enabled by computational workload flexibility remains insufficiently quantified. This paper develops a two-stage electricity--computation co-optimization framework to quantify the individual and coordinated hosting capacity values of spatial and temporal workload flexibility. Data center capacities are determined endogenously, with computational task arrivals scaling with admitted capacity. Spatial routing of eligible interactive requests and temporal deferral of batch workloads are linked to server operation and facility power demand, subject to workload completion requirements and transmission constraints. The first stage maximizes aggregate hosting capacity; the second fixes the resulting site capacities and optimizes annual workload scheduling and generation dispatch to minimize generation and carbon costs. An 8760-hour case study with three data centers on a modified IEEE 118-bus system shows that temporal and spatial flexibility individually increase hosting capacity by 21.26\% and 5.37\%, respectively. Their coordinated use raises hosting capacity from 4635.1~MW to 6231.5~MW, a 34.44\% increase without network reinforcement. The coordinated gain exceeds the sum of the individual gains by 361.9~MW, demonstrating complementarity between spatial migration and temporal deferral. By redistributing computational demand across locations and operating hours, these mechanisms enable fuller utilization of existing transmission capacity. The findings establish workload flexibility as a quantifiable planning resource for accommodating additional data center capacity within existing transmission networks.