PJM Interconnection Queue Reaches 3,500 Data Centre Applications Totaling 95 GW

The PJM Interconnection, which operates the electrical grid across 13 US states and the District of Columbia serving 65 million people, has disclosed that its generation interconnection queue now contains over 3,500 applications from data centre developers requesting a combined 95 gigawatts of new service. The figure is staggering - PJM's total installed generation capacity is approximately 185 GW, meaning the data centre queue alone represents over half of the entire existing grid. No other category of electricity consumer has ever produced demand of this magnitude in the history of the US power system.

The geographic concentration of this demand is equally remarkable. Northern Virginia, which sits within PJM's Dominion Energy zone, accounts for approximately 4,200 MW of existing data centre load and an estimated 3,100 MW of pipeline capacity. Data centre load in the PJM territory is growing at approximately 30% annually - a rate that far exceeds the utility system's ability to add generation and transmission capacity. Dominion Energy's $9.6 billion grid investment plan, announced in 2025, is designed to add 10 GW of new capacity across Virginia by 2035, but even this massive programme cannot keep pace with demand growth.

PJM officials have acknowledged that the vast majority of the 3,500 queue applications will never proceed to construction. Many represent speculative filings by developers seeking to preserve optionality across multiple potential sites - a common tactic in land-constrained markets where securing a queue position years in advance is a strategic necessity. Nevertheless, the sheer volume has overwhelmed the queue management process, with new applicants facing study timelines of 4-5 years before receiving interconnection agreements. The interconnection study process itself has become a bottleneck: PJM's engineering staff must evaluate each application's impact on the grid, and the complexity increases exponentially as more applications interact with the same transmission corridors.

The bottleneck has prompted federal action. The Federal Energy Regulatory Commission (FERC) issued Order 2023-B in early 2026, implementing financial readiness requirements including non-refundable deposits of $5,000 per MW for queue applications. For a 100 MW data centre, this means a $500,000 upfront deposit - enough to deter purely speculative filings while remaining manageable for serious developers. The order also establishes a "fast track" process for projects that can demonstrate site control, utility commitment letters, and engineering readiness, potentially reducing study timelines from 4-5 years to 18-24 months for qualified applicants. Industry observers expect the reforms to clear roughly 60% of speculative applications from the queue within 18 months.

The grid constraints are having profound effects on the data centre market. Developers are increasingly pursuing behind-the-meter power solutions that bypass the interconnection queue entirely. Constellation Energy's 2 GW nuclear PPA, Amazon's agreement with Talen Energy at the Susquehanna plant, and Microsoft's multiple nuclear partnerships all reflect a strategy of securing dedicated generation capacity rather than waiting for grid connections. On-site natural gas generation, fuel cells, and battery storage are also being deployed as interim power solutions while grid connections are processed.

The PJM situation is also driving geographic diversification of data centre development. Markets outside the PJM territory - including Texas (ERCOT), the Southeast (Southern Company territory), and the Midwest (MISO) - are attracting developers who would otherwise have built in Virginia or the mid-Atlantic. States like Mississippi, Louisiana, Indiana, and Wisconsin have become beneficiaries of this displacement, attracting multi-billion-dollar investments from hyperscalers that might have gone to Virginia a few years ago.

For the broader economy, the PJM grid situation raises uncomfortable questions about whether the US power system can support the scale of AI infrastructure that technology companies are planning. The International Energy Agency estimates that data centres globally will consume 1,000 TWh of electricity annually by 2030 - roughly equivalent to Japan's entire consumption. Meeting this demand while simultaneously decarbonising the grid and maintaining affordability for residential customers represents one of the most significant infrastructure challenges of the next decade. The resolution will likely involve a combination of grid expansion, behind-the-meter generation, demand flexibility programmes, and potentially difficult trade-offs between AI growth and other societal priorities.