US House backs grid cost bill targeting AI data centers

For the first time, US federal lawmakers have aimed a major piece of energy policy directly at AI data centers. The Ratepayer Protection Act does not cap AI build‑out, but it rewires who pays for the extra power lines, substations, and generation required to keep hyperscale campuses online. Making 100‑megawatt‑plus sites cover their own grid upgrades will push operators like Microsoft, Amazon, Google, and Meta to sharpen site selection and long‑term power contracts, rather than assuming that local residents will silently subsidize their growth.

In the race to AGI, this is an early sign that physical constraints around power and infrastructure are turning into political constraints. The more communities rebel against rising bills, the harder it becomes for labs to scale frontier training clusters and low‑latency inference regions on autopilot. Over the next few years, we should expect marginal AI compute to get more expensive in jurisdictions that adopt the standard, nudging big players toward regions with friendlier tariffs or abundant renewables. That does not stop AGI work, but it likely redirects future superclusters to places with stronger political support and more patient capital, while raising the bar for smaller challengers that lack deep energy‑market expertise.