A year-by-year balance of new firm generation (nameplate additions × capacity accreditation, net of retirements) against the data-center load that utilities say is in their queues. Each input is a slider. Change them to see whether the queue can be served, then scroll down to see where the real load lands.
Gas additions from 2028 = global heavy-duty turbine deliveries × share landing here × 85% reaching COD, lagged by the delivery-to-COD years, plus the fast lane; 2026–27 use EIA plans. Servable = (cumulative net firm additions − firm needed for non-DC growth) ÷ ((1 + reserve margin) × peak coincidence × (1 − flexible share)). Stated = queue tranches energized on their requested year at full load. Realistic = queue × unique share × realization, delayed, ramped on an S-curve.
The scenario above says how much data-center load is real by 2030. This section asks where it lands. Some of it stays near where it was queued. The rest goes to the states that score best on time to energize, price, tariff certainty, policy pauses, on-site generation, headroom, flexibility products and tax treatment, all of which utilities, regulators and developers can change. Move the levers and the map moves. Click a state to see what it could do.
Albers USA projection (US Census boundaries via us-atlas). Hover for values; click to select. Labels mark the ten largest values.
Allocation = inertia × queue share + (1 − inertia) × mobile share, where mobile share ∝ exp(Σ weight × signal) × queue sharegravity. Signals are standardized across states (speed = −months to energize; price = −industrial rate; on-site gas = log GW announced) or categorical (tariff approved / pending / none; pause / partial / none; headroom; flexibility program; tax exemption). The state attributes are sourced. The weights are judgment. Read the range of outcomes across scenarios, not any one map.
If a utility wins a slice of this load, what happens to everyone else's bill, and how much of the benefit can go back to the developer? Pick the archetype closest to your utility, then adjust.