FERC Tells Grid Operators: Make Room for Flexible AI Loads
Get the Energy newsletter
Daily energy & climate — solar, EVs, oil, the policy fights and tech bets shaping the transition. Free.
- AI data centers and electrification are driving the fastest U.S. electricity demand growth in a generation, with a single campus sometimes requesting more power than a mid-sized city — or, in some cases, than an entire state.
- FERC in June directed regional grid operators to revisit rules for connecting large new loads and to make room for customers that can raise and lower demand on request.
- A University of Utah study modeled the Western grid and found running data centers flexibly could save roughly $62 million a year through off-peak scheduling alone, rising to as much as $590 million with regional coordination and on-site energy infrastructure.
- University of Utah researcher Mohammad Amin Mirzaei summed up the logic: "Instead of shipping electricity across overloaded lines, you ship the computation."
- Torus pairs flywheel storage — millisecond response built on centuries-old rotational inertia principles — with battery storage that handles minutes-to-hours events, producing a hybrid resource that holds voltage and frequency steady at substations or across enrolled customer sites.
- Torus builds its systems in Utah and goes from signed contract to operation in 12 to 16 weeks, offering utilities a capacity option on a timeline far shorter than multi-year grid expansion.
- Retiring conventional power plants are removing the rotating-mass "shock absorber" that absorbed sudden frequency swings, shrinking the cushion the grid relies on as more power flows through electronics instead of spinning machines.
Why it matters: For utilities facing interconnection queues that outrun construction timelines by years, FERC's June directive reframes load flexibility from a niche offering into a grid-planning requirement. The Utah study's $590 million savings ceiling turns flexible data center operation into a documented system-level value, shifting the burden onto hyperscalers to design for dispatchability.
