Future data centers are driving up forecasts for energy demand (apnews.com)

U.S. utilities and grid operators are projecting eye-popping electricity demand increases—often two to three times current levels within a few years—driven largely by massive new data centers supporting AI workloads. That surge has triggered scrutiny from state lawmakers, regulators and FERC, who warn forecasts may be inflated by speculative or duplicate interconnection requests (developers sometimes file with multiple utilities and don’t disclose overlaps). Regulators fear ratepayers could be forced to underwrite unnecessary generation and grid builds, while grid operators confront tight timelines to add capacity for reliability. Technically, the problem is forecasting and commercial-readiness verification: PJM and ERCOT flagged that single projects can be counted multiple times across territories, and utilities often lack standardized vetting to confirm customers have financing, contracts or committed clients. Some utilities, like PPL, say projects are backed by sizable financial commitments and will materially raise peak demand by 2030, while others point to examples where customer bills have already risen (PECO) due to data-center-driven wholesale costs. Responses so far include a FERC information request, Texas laws forcing disclosure of duplicate requests and proof of financial commitment, and industry calls for best practices. The takeaway for AI/ML stakeholders: accelerated compute growth demands better cross-jurisdictional data, transparent commercial verification, and coordinated planning to avoid stranded assets and preserve grid reliability.