The split between who is driving demand and who is paying for it is the clearest fingerprint AI has left on US power. On a 12-month rolling-sum basis, commercial electricity sales were 12.4% above January 2022 by March 2026, while residential sales rose only 2.3%. In the latest reading, commercial sales grew 2.6% year over year while residential sales fell 0.2% — a 2.9-percentage-point gap. Yet residential prices climbed nearly twice as fast as commercial prices over the same window.
The mechanism is grid economics. Capacity additions, transmission upgrades, and reserve procurement get spread across the rate base, and residential customers are the largest, least mobile slice of that base. When a utility builds to serve a data center that has signed an interconnection agreement, the fixed costs of that buildout often land on every ratepayer, not just the load that triggered them. Consumer Reports cited a 10.2% average residential-rate increase between March 2025 and March 2026, with AI and data centers named as contributors.
Rate cases across the country now reference data-center load explicitly. Utilities asked regulators for a record $31 billion in additional cost recovery tied to powering AI data centers. In New Jersey, the governor laid out a plan specifically to ease the burden AI data centers place on towns and electric bills. The political response is catching up to a cost-allocation problem that the price data already shows: households are subsidizing a load shock they did not create.
This is not proof that AI alone caused the increases — fuel costs, storm hardening, and inflation all push rates up. But the timing, the commercial-residential demand split, and the explicit data-center references in rate filings line up with the cost-socialization story rather than against it.
