The biggest U.S. power grid is under strain from AI — and no one is happy

PJM Interconnection says its system has "years, not decades" to undergo fundamental changes as AI and cloud-driven electricity demand strains its grid and interconnection process. The operator paused new generation applications in 2022 amid a backlog; of more than 300 GW in the queue, only 103 GW signed agreements and just 23 GW have connected so far. The issue is becoming material for data centers and power companies, with AEP considering exiting PJM altogether.

Analysis

The market is underestimating how quickly grid bottlenecks can translate into a capital allocation regime change. PJM’s failure is not just a utility-planning issue; it is becoming a gating factor for AI infrastructure deployment, and that shifts bargaining power from hyperscalers to regulated utilities, transmission owners, and developers with already-approved queue positions. In the near term, the scarcity premium will likely accrue to firms that can monetize existing interconnection rights rather than those dependent on new greenfield access.

AEP’s threat to exit PJM matters less as an operational event than as a signaling device: if one large incumbent is openly questioning the platform, the probability of more aggressive governance reform rises, but so does the risk of fragmentation and legal delay. That creates a multi-year overhang for marginal generation projects in the region, while also supporting higher implied capacity prices and transmission investment elsewhere. The second-order effect is that capital may rotate away from pure merchant development and toward balance-sheet-heavy utilities and IPPs with regulated or contracted cash flows.

The most important contrarian point is that the queue backlog may be more bullish for incumbents than bearish, because it slows supply response precisely when demand is inflecting upward. That implies the pricing power improvement in regional power markets could persist longer than consensus expects, especially if AI load growth remains elastic only at the margin. The tail risk is policy intervention: a forced process overhaul, faster permitting, or federal pressure on PJM could compress scarcity rents within 6-12 months, but absent that, the bottleneck likely remains a 2-3 year earnings tailwind for owners of existing assets.

Key Decisions for Investors

• Long AEP on any post-earnings weakness over the next 1-2 weeks; the market is likely over-discounting governance risk while underpricing the value of a more regulated, less congested footprint. Target a 10-15% relative outperformance versus the utility index if PJM reform drags out.
• Pair trade: long NRG / short merchant-exposed developers with heavy PJM new-build dependence over 3-6 months. The thesis is that constrained interconnection preserves scarcity pricing for existing capacity while punishing names needing fast queue access.
• Buy 6-12 month calls on transmission-oriented utilities or infrastructure owners with capital projects already in-service or near in-service. Use strikes 10-15% OTM to express upside from higher allowed returns and increased capex urgency.
• Short-dated bearish options on pure-play AI data center infrastructure providers reliant on PJM expansion approvals, if available, as a hedge against queue delays. The risk/reward improves if the stock has priced in near-term power availability assumptions.
• Reduce exposure to speculative generation developers without secured interconnection positions; the bottleneck creates a winner-take-most dynamic where late entrants face multi-year delays and higher financing costs.