Peak Demand Growth

US electricity demand was flat for 15 years. It is now growing again, and the new growth is coming from sources that concentrate load in ways the grid was not designed for.

Three demand drivers are converging simultaneously. Data centers (AI training clusters consuming 50-100 MW each, with some facilities approaching 500 MW). Electric vehicle charging (a fleet of 50 million EVs would add roughly 50-70 GW of potential peak demand if unmanaged). Building electrification (heat pumps replacing gas furnaces shift winter heating load onto the electric grid, potentially creating new winter peaks in regions that historically peaked in summer).

The grid concern is not total energy consumption (which grows gradually) but peak demand concentration. A 100 MW data center is a single point load larger than many substations can serve. Fifteen Level 2 EV chargers on a single residential transformer can overload it. Heat pump adoption in a cold region can flip the annual peak from summer to winter within a few years.

The data center surge. ERCOT's most aggressive forecasts, built from interconnection requests that will not all materialize, put possible data center load in Texas at 30-50 GW by 2030, roughly half of ERCOT's current peak demand.

If demand growth is concentrated in data centers, can it be managed differently than residential growth?

Yes, because data centers are flexible. Unlike residential demand (dispersed, price-insensitive), data centers can negotiate interruptible tariffs, co-locate with generation, and shift computational workloads to off-peak hours. The policy question is whether utilities and regulators will require them to.

---