America is building datacenters faster than the grid can power them

Moody's projects US datacenter electricity demand will nearly double to 426TWh by 2030, requiring about $110B of additional generation capacity and adding $25B-$30B annually to electricity-system costs. Grid interconnection, permitting and equipment bottlenecks can delay projects by up to seven years, with Jefferies estimating that roughly half of planned 2026 US datacenter capacity will not come online on schedule. Rising local power bills are driving political and community opposition, including tighter approval rules and moratoriums, although Moody's expects long-term generation and transmission investment to improve grid reliability and affordability.

Analysis

The binding constraint is shifting from GPU availability to energized megawatts, which changes the AI capex value chain. Utilities with rate-base growth, transmission exposure, and constructive regulatory regimes—particularly CEG, VST, NRG, PWR and ETN—should capture a larger share of AI infrastructure spend than data-center landlords until interconnection queues clear. The more consequential bottleneck is electrical equipment: multi-year lead times for transformers, switchgear and high-voltage cabling can sustain pricing and margins for ETN, HUBB and PWR even if announced data-center projects slip.

For hyperscalers, delayed energization converts nominal capex into lower-return construction-in-progress and defers revenue recognition from AI services. That is most material for smaller, highly levered GPU-cloud operators and speculative data-center developers, where financing assumptions often require rapid capacity turn-up; it is less damaging to MSFT, GOOGL, AMZN and META, which can redirect workloads across existing regions. In the next 1-3 months, interconnection-delay disclosures and utility load-forecast revisions are likely to matter more for AI-adjacent equities than another incremental model launch.

The consensus may be overpricing nationwide power scarcity while underpricing geographic dispersion. Texas and other faster-permitting markets can attract disproportionate workload and industrial investment, but local affordability backlash creates a regulatory call option against merchant generators and developers relying on unconstrained retail-rate pass-through. Over 6-18 months, dedicated generation and customer-funded transmission should reduce the scarcity premium; the best structural longs are therefore equipment and engineering suppliers rather than pure power-price beta. The thesis is falsified if utility capital plans flatten, transformer lead times normalize materially, or hyperscalers cut 2026-27 data-center capex rather than shift project geography.

Key Decisions for Investors

• Initiate a 6-12 month long ETN / short EQIX pair: ETN benefits from electrical-content inflation regardless of which campuses receive power, while EQIX faces a higher risk that development yields are delayed by interconnection timing. Target 15-20% relative upside; exit if ETN backlog growth decelerates below low-double-digits or EQIX raises development-return guidance despite grid delays.
• Accumulate PWR and HUBB on market weakness for a 12-18 month infrastructure cycle, sized as lower-beta AI exposure. Risk/reward improves if transmission approvals convert into awarded projects; cut exposure on evidence of utility capex-plan deferrals or a broad reversal in transformer/switchgear lead times.
• Maintain selective long CEG or VST only against a hedge in XLU: firm-power scarcity supports realized pricing, but this trade should be tactical over 3-6 months because ratepayer intervention and new capacity can cap returns. Take profits if regulatory filings restrict data-center cost recovery or forward power spreads compress sharply.
• Avoid adding to leveraged private-equity-style data-center developers until project-level evidence shows contracted power, not merely land and customer demand. Create an alert around hyperscaler earnings: any material reduction in 2026 capital plans would pressure the entire power-equipment chain, including ETN, HUBB and PWR.