PJM trims summer 2027 peak demand forecast to ~160 GW from ~164 GW because some projects, including data centers, lack firm service or construction commitments
The biggest US grid operator dialed back its forecast for power demand growth, offering a reality check to the frenzy around the artificial intelligence boom.
Context & Ripple Effects
PJM had just projected 4.8% average annual demand growth over the next decade, making the quality and timing of proposed large loads central to its planning assumptions. This revision distinguishes projected demand from projects that have secured service and construction commitments.
The adjustment matters in a region already facing projected electricity-bill increases tied to grid strain, because speculative or incomplete load requests can distort the capacity outlook used by grid planners and customers.
First-order effects
- PJM lowers the peak-demand assumption it will use for summer 2027 planning, removing roughly 4 GW of projects that lack firm service or construction commitments from the near-term outlook.
- Data-center and other large-load developers without those commitments face a clearer signal that proposed capacity will not automatically be treated as dependable demand in PJM forecasts.
Second-order effects
- The lower forecast can reduce immediate pressure on PJM's supply planning relative to the prior estimate, while concentrating attention on projects that can demonstrate credible interconnection and build timelines.
- Large data-center developers may be pushed toward firmer power arrangements or flexibility commitments, consistent with PJM's later proposal that major facilities provide generation or curtail load to protect grid reliability.
Third-order effects
- If grid operators increasingly screen demand forecasts for execution certainty, AI-infrastructure planning will shift from headline project pipelines toward contracted power, interconnection status, and construction readiness.
- That would make electricity availability a more explicit gating factor for compute expansion, with demand forecasts becoming more volatile as projects advance or fail to secure service.
The trend: AI-driven power demand is moving from broad growth projections toward a more disciplined test of which data-center projects can convert planned compute into deliverable grid load.