Regulators, not markets, decide who pays for data-center electricity
A distinct large-load tariff class with minimum-take obligations, long contract terms, and exit liabilities becomes the standard treatment across the…
Claude · 2027 · plausible
Prior state
Data-center load growth has pushed mid-Atlantic capacity prices to their administrative caps, retail bills have risen visibly in several states, interconnection queues are multi-year, new gas turbines are sold out for years, and one large utility settlement has already created a distinct rate class for very large customers with minimum-take and exit terms. Federal proceedings on co-located generation remain unresolved.
Material change
A distinct large-load tariff class with minimum-take obligations, long contract terms, and exit liabilities becomes the standard treatment across the major affected US markets, and siting moves from a purely local permitting matter to a state legislative one. The default cost incidence of interconnection and capacity shifts from the general ratepayer to the hyperscale customer, and in several states new load is made conditional on bringing its own generation.
Why now
Rate cases filed in 2025 and 2026 reach decision in 2027; the capacity auction results already cleared feed into retail bills during the year, making the cost politically visible; and state legislatures seated after the 2026 elections take up the siting and tariff bills in their first sessions. One long-planned reactor restart is scheduled to return to service in the same year, which sharpens the argument about whether new load can procure its own supply.
Mechanism and resistance
Public utility commissions set tariffs through adjudicated proceedings in which utilities, large customers, and consumer advocates all participate; legislatures set siting and tax-abatement rules. Resistance comes from utilities whose earnings grow with rate base, from states competing for investment, from the federal regulator's jurisdiction over wholesale arrangements, and from the difficulty of writing minimum-take terms that survive a demand slowdown.
Consequences
Residential and small-commercial customers in the affected states are the intended beneficiaries; the compute buildout's cost of capital rises and its geography shifts toward jurisdictions with spare generation, cheap gas, or permissive siting — including outside the US. Utilities gain contracted revenue and a stronger investment case for new generation; the political salience of electricity prices, which has already reordered several state races, becomes a fixed feature.
End state
The US enters 2028 with large-load tariffs as the norm in the major data-center states, and with the cost of new compute capacity explicitly priced rather than socialised.
Observable test
The number of states with approved large-load tariff classes at year-end, the volume of signed minimum-take contracts, and whether residential rate increases in the mid-Atlantic decoupled from data-center load growth.
Disconfirming sign
Commissions continuing to socialise interconnection and capacity costs across the general ratepayer base, or a capex slowdown that removes the pressure before the cases are decided.