Dynamic AI simulation to
guide
strategic decision making
UTILITY
HYPERSCALER
GOVERNMENT
INSTITUTIONALOn a hot Amarillo afternoon, an unprotected turbine fleet loses output exactly when AI tenants need it most. Built on Crusoe's own turbine model data, this simulation plays a design hot day hour by hour — and prices the recovered capacity as contracted AI compute.
Output falls with every degree of afternoon heat. The capacity Crusoe can commit to tenants is capped by this worst hour — not by the nameplate of the fleet.
Chilled inlet air removes the hot-day derate: the fleet delivers 734 MW at the peak hour regardless of ambient — capacity that can be contracted firm, year-round.
Ice built overnight carries the chilling burden through the peak window, so no recovered megawatt is spent running chillers at the hour it is worth most: 787 MW, firm.
The block of fleet output that would otherwise run chillers at the peak steps off in minutes when the ice discharges — without touching a single rack. The same asset that holds 787 MW firm is a controllable flexibility lever at the hours the grid is shortest.
Texas SB6 co-location rules now attach a 30-minute grid-emergency curtailment obligation to campuses like Goodnight. Ice discharge is a fast, verifiable response lever inside that window: chilling demand steps off the fleet on dispatch, freeing generation toward the emergency — while compute keeps running.
Scarcity pricing peaks on the same hot afternoons that derate an unprotected fleet. Every megawatt of chilling the ice carries through the discharge window is a megawatt of protected generation freed to flow across the meter at scarcity-hour prices, where interconnection allows.
