The rapid expansion of hyperscale data centers is placing growing pressure on electricity systems, land, and climate-sensitive infrastructure alike. While maps of existing facilities are now available, a more fundamental question remains unresolved: where can additional hyperscale data centers still be built under present-day physical and infrastructural constraints?

In a new arXiv preprint, I approach this question using a constraint-first, national-scale geospatial framework using a dozen nation-wide high-resolution geospatial data set covering the US. Rather than forecasting demand or optimizing sites, the analysis infers feasibility from revealed hyperscale siting behavior, conditioned on power-grid adjacency, climate and cooling constraints, flood risk, land use, terrain, and regulatory proxies.

Press enter or click to view image in full size

Two independent, unsupervised spatial methods — both calibrated exclusively on existing hyperscale environments — converge on a limited feasible land envelope. Translating this envelope using observed hyperscale power densities suggests that total physically feasible U.S. hyperscale capacity lies in the tens of gigawatts rather than the hundreds, under current infrastructure and siting patterns.

These estimates should be interpreted as physical and infrastructural ceilings, not forecasts of construction or investment. Substantially exceeding these bounds would likely require structural changes to the energy system, regulatory environment, or prevailing data center siting practices.

Full paper (arXiv): https://arxiv.org/pdf/2602.02529