AI data centers are outgrowing the grid. Hyperscale campuses need hundreds of megawatts on tight timelines, but transmission upgrades, substation builds, and utility approvals can take most of a decade. To keep schedules intact, developers are installing on-site aeroderivative gas turbines—compact, fast-start units originally designed from jet engines—to generate power immediately. In several projects, these systems are serving as the primary power source until (or unless) full grid service arrives.

This breaks the long-standing assumption that data centers must wait for centralized grid capacity. Control over power now determines time-to-revenue. On-site generation compresses schedules, de-risks capital deployment, and shifts leverage away from utilities. The trade-off is higher near-term capex and fuel exposure, but for AI operators, delay costs more than diesel or gas.

Expect regulators to face pressure to reconcile emissions policy with reliability reality. Hybrid models—on-site turbines paired with grid tie-ins, storage, and future cleaner fuels—will become standard. Utilities may respond with expedited programs, but developers won’t wait. Power strategy is becoming site strategy.

In AI infrastructure, power access is the critical path. Whoever controls megawatts controls deployment—and increasingly, that means building the power plant first, not waiting for the grid.