Behind-the-Meter Solar at Hyperscale Data-Center Campuses: A Resource-Graded Carbon-Price Threshold Across Eleven U.S. Markets

Abstract

Grid interconnection and delivery capacity, rather than generation cost, have become the binding constraint on serving hyperscale electricity loads in many U.S. markets. We show that the carbon price required to bring on-site solar into a cost-optimal campus portfolio can be read directly from the local solar resource: across eleven U.S. data-center market archetypes the threshold falls from approximately $230/ton at an annual solar capacity factor of 0.147 to $110/ton at 0.238, a near-linear relationship (R²=0.97) that gives developers a single-parameter screening rule for siting. The result comes from a reproducible framework built only from public data, coupling a levelized-cost and Monte Carlo screening layer to a temporally-resolved capacity-expansion and dispatch optimization that co-optimizes investment and hourly operation and prices battery storage explicitly, run on real hourly NREL PVWatts V8 profiles. Because dispatchable gas capacity is retained for firmness in every scenario, on-site solar competes against gas’s variable cost rather than its full levelized cost; this is why the threshold is graded by resource and why it lies well above current U.S. carbon prices, with gas remaining least-cost firm supply in every market at carbon prices up to at least $100/ton. The relationship survives more favorable PV assumptions and a temperature-responsive cooling load. All code, inputs, and outputs are openly archived.