AI Data Centers Are Now Driving Your Electric Bill — and State Utility Commissions Are Where the Real AI Policy Fight Is Happening

The Grid-Connection Backlog Is Forcing Utilities to Build Faster Than They Can Plan

The scale of the mismatch driving residential rate increases is visible in the interconnection request data: utilities across the country received requests for at least 700 gigawatts of new power connections in 2025 alone, more than the entire United States consumed across all sectors in 2023. Most of these requested projects will never be built — developers routinely file speculative interconnection requests at multiple sites to preserve optionality while they finalize land, power purchase agreements, and financing — but utilities cannot easily distinguish real demand from "phantom load" at the planning stage, and the cost of building generation, transmission, and substation capacity to accommodate even a fraction of these requests is being recovered through rate cases that ultimately land on residential and small business ratepayers. Microsoft's disclosure of an $80 billion backlog of Azure orders it cannot fulfill due to power constraints is the clearest evidence that the demand is not phantom in the aggregate, even if individual project requests are oversized — it is a genuine supply problem, with the average grid-connection wait time in primary data center markets now exceeding four years, meaning a project breaking ground today may not receive a usable utility interconnection until 2029 or 2030.

Virginia's State Corporation Commission response — approving a new GS-5 rate class in November 2025 specifically for data center customers consuming over 25 megawatts, requiring them to pay for at least 85 percent of their contracted distribution and transmission demand starting January 2027 — is the first serious regulatory attempt to make data center operators internalize the grid upgrade costs their demand is driving, rather than spreading those costs across the general ratepayer base. The PJM Interconnection capacity auction price increase of 833 percent for the 2025-2026 delivery year, directly attributable to data center demand growth outpacing generation additions in the mid-Atlantic region, is the wholesale market signal that state regulators across other data-center-heavy states — Texas, Georgia, Arizona, Ohio — are now studying as a template, because the alternative to a dedicated large-load rate class is residential customers in those states experiencing the same cost socialization Virginia ratepayers absorbed before the GS-5 class existed.

The Reshoring Story and the Grid Story Are the Same Story, and Most Coverage Misses That

The connection between this electricity rate dynamic and the broader US manufacturing reshoring narrative is closer than most coverage acknowledges. Data centers and power generation infrastructure are, according to industrial construction spending data, the actual locus of US industrial construction growth in 2026 — not the factory reshoring that tariff policy is explicitly designed to incentivize. Manufacturing construction spending, excluding electronics and semiconductor fabs, rose only modestly since the current tariff regime began, while data center and power generation construction spending has grown sharply, meaning that the most visible domestic industrial buildout Americans are actually seeing reflected in their utility bills is AI infrastructure, not the steel mills and assembly plants that tariff policy was sold as protecting. For households experiencing both higher prices on tariffed consumer goods and higher electricity bills from data center-driven grid investment simultaneously, the two policy stories that dominate 2026 economic coverage — trade protectionism and AI infrastructure investment — are landing on the same family's monthly budget from two different directions, and neither was originally framed to the public as a household cost increase.

The third-party solar and battery ownership data — growing toward capturing up to 69 percent of residential installations in 2026, up from roughly 45 percent — is the household-level adaptation response to a rate environment that state regulators have not yet fully stabilized, and it is concentrated specifically in the geographic markets where data center buildout is heaviest. This is creating a feedback dynamic state utility commissions are only beginning to grapple with: as data center load drives rates up, more affluent or proactive homeowners shift toward behind-the-meter generation and storage, reducing their grid dependence and their contribution to the fixed-cost recovery that utilities need from a broad ratepayer base, which in turn increases the rate pressure on customers who cannot afford or access solar and battery alternatives. The households least able to install solar — renters, lower-income homeowners, and those in older housing stock — are increasingly the ones bearing a disproportionate and rising share of grid infrastructure costs that AI data center demand created.