Power prices are up 76% on America’s biggest grid, and a watchdog is pointing fingers

Wholesale electricity prices in PJM nearly doubled year over year to $136.53 per megawatt-hour from $77.78, with the independent monitor blaming surging data center demand and PJM’s failure to adapt. Monitoring Analytics warned the price impacts on customers are already large and not reversible, and likely to worsen near term if data center load is not addressed. The article highlights a structural supply-demand problem in the U.S. power grid tied to AI-driven electricity demand, with implications for utilities, power generators, and data center operators across the PJM region.

Analysis

This is a demand-shock story, but the market is still pricing it like a transient congestion issue rather than a structural re-rating of the regional power stack. The key second-order effect is that hyperscaler load is now functionally acting like a tax on every incremental consumer and industrial customer in the zone, which should widen the political spread between “AI buildout” beneficiaries and ratepayers bearing the cost. That tends to favor assets with contracted load growth and penalize utilities with weak pass-through or poor transmission access.

The bigger tell is not spot power prices, but capacity scarcity extending into future auctions: once data centers consume the reserve margin, the system starts rewarding whoever can bring firm capacity online fastest, not whoever has the cheapest energy. That changes the relative economics of gas peakers, grid equipment, interconnection software, and transmission bottleneck relief versus long-duration renewables that still need firming. Expect a multi-quarter reallocation of capex toward reliability, with regulators forced to choose between approving higher bills or delaying AI interconnects.

The contrarian view is that the move may be under-discounted in utilities and over-discounted in infrastructure suppliers. If the load growth is real and sticky, the winning trade is not just higher power prices, but a sustained upgrade cycle in transformers, switchgear, substations, and grid software with backlog visibility. Meanwhile, data center operators face a hidden margin squeeze if power becomes the gating input; the market has largely treated compute as the constraint, but power availability may become the binding variable first.

Tail risk is regulatory intervention that caps pricing power or accelerates rule changes on large-load interconnection within 3-9 months, which would soften the near-term bull case for power merchants but not fully reverse the infrastructure spending cycle. The reverse catalyst would be a faster-than-expected wave of new gas and transmission approvals, which could relieve 2026-27 scarcity, but that is a years-long process. Near term, the trade remains a spread between scarce firm capacity and everything exposed to unpriced load growth.

Key Decisions for Investors

• Long CEG / short XLU for the next 3-6 months: own the merchant/firm-capacity exposure while fading regulated utilities that cannot fully reprice or re-rate quickly; target a 10-15% relative move if capacity tightness persists into the next auction cycle.
• Long ETN and HUBB on 6-12 month horizon: backlog expansion in switchgear, transformers, and electrical distribution should be one of the cleanest second-order beneficiaries; risk/reward improves on any pullback toward 15-20x forward earnings.
• Long NEE Jan-2026 calls, but only as a volatility expression: if PJM-style scarcity forces more transmission and firming investment, utility quality should re-rate, yet the equity remains hostage to rate politics; use defined risk.
• Short data-center-adjacent REITs with poor power access economics on strength over the next 1-2 quarters; prefer names with high lease concentration in Northern Virginia-style markets where power is increasingly the bottleneck rather than demand.
• Pair trade: long PWR / short AI-heavy software basket if markets begin pricing accelerated grid buildout; infrastructure execution wins before end-user AI monetization does, and the spread should widen over the next 6-9 months.