Introduction

November 15, 2025 — Facing mounting strain on electricity systems and public concern over rising rates, policymakers in multiple U.S. states are debating whether to curtail or disconnect large data centers during grid emergencies. In Indiana, a new grassroots campaign has rallied bipartisan support for a temporary moratorium and rate redesign to protect residential customers. Elsewhere, lawmakers have floated “tough-love” measures that would bump data centers off the grid when capacity is scarce—an unmistakable sign that AI’s appetite for power has become a political issue (see reporting today from Michigan Advance and Milwaukee Independent). (Michigan Advance)

Why it matters now
• Policy inflection: emergency cutoffs would put compute behind hospitals and homes on priority lists.
• Cost pressure: proposals target subsidized tariffs and opaque incentives that shift risk to ratepayers.
• Siting squeeze: opposition complicates speed-to-power strategies for hyperscalers and colocation providers.
• Signal to markets: reliable power access—not land—is emerging as the new gating factor for AI growth.

Call-out
Power is the new platform risk for AI—and policymakers know it.

Business implications
For hyperscalers, the message is stark: capacity planning must assume intermittent grid access in certain markets. That pushes designs toward on-site and near-site resources—such as long-duration batteries, thermal storage, and firm contracts with independent power producers—and toward software adaptations, including workload shedding, job preemption, and model-aware scheduling during scarcity windows. Developers will need earlier and deeper engagement with utilities and regulators, as well as community benefit agreements that are specific, measurable, and enforceable.
Enterprises will see regional divergence. In cooperative jurisdictions, AI services should scale on schedule but arrive with new power-aware SLAs; in contested regions, expect quota systems, time-of-use pricing, and “bring-your-own-power” expectations that favor organizations with energy procurement expertise. Equipment makers—from transformers to switchgear to power electronics—face demand upside but must design for faster interconnection and island-mode operation.

Looking ahead
Near term (3–9 months): expect emergency-curtailment bills in several legislatures, along with transparency rules around incentives and special rates. Utilities will publish clearer scarcity protocols; cloud providers will pilot workload failover tied to ISO/RTO alerts.
Longer term (12–36 months): campus-level microgrids, behind-the-meter renewables, and firming assets become standard line items. Power procurement becomes a core competency for AI leaders; “usable compute per kWh” emerges as a board-level metric. Projects in contested regions adopt hybrid models—part grid-tied, part self-powered—to ensure service continuity.

The upshot
The AI boom is colliding with grid reality. Winners will not simply buy more GPUs—they will design for energy scarcity, earn a social license through credible community benefits, and treat resilient power as an integral part of product quality. In a world of rolling constraints, the strategic edge belongs to those who can keep the lights—and the models—on.

References
• Michigan Advance — “A New Unifying Issue: Just About Everyone Hates Data Centers,” Nov 15, 2025. (Michigan Advance)
• Milwaukee Independent — “Big Tech faces pushback as lawmakers consider kicking energy-hungry data centers off U.S. power grids,” Nov 15, 2025. (Milwaukee Independent)
• Stateline — “Most states don’t disclose which companies get data center incentives, report finds,” Nov 12, 2025. (Stateline)