Data-center tariff politics

The article links AI data-center demand to a private equity push into utilities and warns of rate hikes, directly engaging the question of who pays for grid upgrades; PE ownership could pressure regulators to shift costs from hyperscalers onto general ratepayers unless bespoke tariffs are enforced.

The article’s $14.5B 'Project Marvel' hyperscale campus (18 buildings on ~100 acres of clear-cut within a 700-acre site) highlights a resource-hungry load in a regulated-utility state; residents fear power/water strain while 'state and local governments compete' for projects—conditions that trigger bespoke utility rate design and cost-allocation fights central to this idea.

Epoch’s forecast on 2030 frontier AI training power demand implies major grid upgrades and cost allocation choices; it strengthens the case for bespoke rate design so hyperscalers fund buildouts rather than general ratepayers.

The article argues deep-pocketed AI companies could overbid scarce capacity and raise bills unless policy forces them to pay for new infrastructure that also benefits others.

OpenAI–Oracle’s 4.5 GW Stargate expansion toward a 10 GW buildout and the plan to feed clusters that swallow gigawatts force regulators to design bespoke rates and cost-allocation so hyperscalers fund grid upgrades without burdening ratepayers.

Zuckerberg’s plan for several multi‑GW clusters (Prometheus in ’26; Hyperion up to 5 GW) implies massive grid upgrades and bespoke cost allocation, exactly the regulatory question of how to prevent hyperscaler costs from being shifted onto general ratepayers.