US Senate rejects bill targeting AI data centre electricity costs

The US Senate rejected the Ratepayer Protection Act, 57-43, short of the 60 votes needed to overcome a filibuster; the bill would have required state regulators to consider making data centres and other major power users bear infrastructure costs. The defeat leaves no new federal framework to shield households from electricity-price and water-supply pressures tied to AI-driven data-centre expansion, despite House passage by 417-3. Senate Democrats opposed the voluntary standard as toothless and advocated mandatory requirements, creating continued regulatory uncertainty for data-centre developers and utilities.

Analysis

The failed federal path shifts the binding constraint from Washington to state public-utility commissions, where outcomes will be materially more heterogeneous. Utilities with concentrated large-load exposure face a widening gap between headline rate-base growth and realized equity returns: incremental transmission and generation capex is valuable only if interconnection deposits, minimum-load commitments, and rider structures keep residential customers insulated. Dominion Energy (D), American Electric Power (AEP), Duke Energy (DUK), and PPL (PPL) should therefore trade increasingly on commission-specific cost-allocation risk rather than a generic AI-power-demand premium.

The cleaner near-term beneficiaries are grid-equipment and engineering vendors—GE Vernova (GEV), Eaton (ETN), and Quanta Services (PWR)—whose order books benefit whether infrastructure is funded by utilities, developers, or hyperscalers. A state mandate requiring large-load customers to pre-fund upgrades would likely reduce data-center project IRRs but would not eliminate the physical need for substations, transformers, transmission, and backup generation; it could actually accelerate procurement by clarifying who pays. Over 6-18 months, merchant power producers with scarce generation near load pockets, notably Vistra (VST) and Constellation Energy (CEG), retain pricing upside, although political scrutiny of retail bills raises the probability of adverse market-design intervention.

Consensus may be over-reading the federal setback as a deregulation win for data-center developers. The greater risk is a patchwork of state-level rules, longer interconnection queues, larger upfront security deposits, and community opposition—raising capital intensity and delaying revenue recognition for Digital Realty (DLR), Equinix (EQIX), and hyperscaler-led builds. The key 1-3 month catalyst is not another Senate vote but utility commission dockets in Virginia, Texas, Ohio, and PJM-related capacity and interconnection decisions; these will establish whether large-load demand is genuinely incremental to utility earnings or becomes a political liability.

Key Decisions for Investors

• Favor a 3-6 month long GEV and/or PWR versus D or AEP pair: equipment and EPC vendors monetize grid buildout regardless of payer, while regulated utilities carry asymmetric exposure to unfavorable cost allocation. Reassess if utilities disclose fully contracted, non-bypassable large-load riders covering incremental capex.
• Maintain a tactical long CEG or VST only on evidence of contracted data-center load or tightening regional reserve margins; use a 10-15% risk budget because retail-rate backlash or capacity-market rule changes can compress scarcity multiples quickly.
• Avoid adding to DLR and EQIX ahead of state interconnection and cost-recovery clarity. Upgrade the group only if disclosed customer contracts demonstrate that power-upgrade deposits and recurring transmission charges are passed through without reducing development yields.
• Set policy alerts for Virginia SCC, Texas PUCT, Ohio PUCO, and PJM proceedings over the next 90 days. A mandatory large-load contribution rule is bearish for data-center development cadence but bullish for GEV/ETN/PWR order visibility; a ruling socializing costs would reverse the utility short leg.