Energization risk: a decision framework for powering data centres, and what it implies for interconnection and tariff policy

Abstract

Data-centre load is being connected to the United States power system faster than that system has historically added firm capacity. In most locations the binding constraint is not a shortage of energy resources but a coordination failure: developers cannot obtain credible dates for when power will arrive, utilities cannot distinguish real load from speculative load in their forecasts, and both sides sign decade-long commitments whose risks neither can quantify. This paper sets out a decision framework that treats energization risk, the joint uncertainty over how much firm capacity a site can obtain, by when, at what cost, and with what exposure if the plan slips, as the object of analysis, and draws out what it implies for the interconnection and tariff rules now being written. An illustrative application produces three results with direct policy consequences: the efficient set of supply plans changes depending on whether plans are judged at the median or the ninetieth percentile; modest correlation among queued requests widens the range of realised capacity by a factor of 2.4 without changing its average; and curtailment commitments buy speed in discrete steps rather than smoothly. The paper argues that FERC's June 2026 large-load orders can create, for the first time, a record capable of calibrating load forecasts, but only if the posting format preserves what was knowable when, and that minimum-take obligations, flexibility tariffs, and cost allocation should each be sized against distributions rather than single figures.