Hyperscale data center growth has become a dominant force reshaping the nation’s largest wholesale power market, driving capacity shortages, inflating auction prices, and forcing a growing debate over whether existing grid structures can support AI-scale infrastructure demand.. A new report from PJM Interconnection’s independent market monitor argues that large-scale data center expansion – particularly forecast AI-related demand growth – now sits at the center of the transmission organization’s escalating capacity market turmoil.. “Data center load growth is the primary reason for recent and expected capacity market conditions, including total forecast load growth, the tight supply and demand balance, and high prices,” the report states.. The report estimates that existing and forecast data center load growth has added more than $23 billion to PJM capacity market costs since 2025. It also argues that without data center demand growth, PJM would not have experienced the same supply-demand imbalance or auction price spikes.. Related:Utilities Say Data Centers Could Lower Electricity Bills. Regulators Want Proof. The warning arrives as PJM, which serves 67 million people across 13 states and Washington, D.C., grapples with a projected capacity shortfall that widened from 208.7 MW in the 2026/2027 Base Residual Auction to more than 6.5 GW in the 2027/2028 auction.. In a December FERC complaint, Gov. Josh Shapiro argued that load growth, interconnection delays, and auction issues led to record power prices. (Photo by Matthew Hatcher/Getty Images). When Power Price Signals Stop Working. The PJM report repeatedly questions the credibility of long-range data center demand projections feeding the organization’s planning models and auction assumptions.. “Forecast data center load growth has been the primary cause to date and the accuracy of those forecasts is highly questionable,” it reads.. That concern extends well beyond the market monitor. In a complaint filed with the Federal Energy Regulatory Commission last December, Pennsylvania Governor Josh Shapiro and the Commonwealth of Pennsylvania argued that significant load growth, the country’s “most snarled interconnection queue,” and a compressed auction schedule had collided to produce record prices that were ineffective at delivering new power generation.. The filing directly tied generative AI and data center expansion to the shift, arguing that “rapidly growing interest in generative AI and associated data centers” had overturned decades of relatively flat load-growth assumptions.. The complaint also identified a contradiction at the center of the current market design. Capacity prices are designed to incentivize new power generation. However, severe congestion in the interconnection system is preventing new supply from responding promptly.. Related:Gridlock or Growth? ERCOT Warns Texas AI Power Boom May Not Materialize. “It is currently physically impossible for new resources to respond to high BRA signals and enter PJM’s marketplace,” the filing states.. More than 3,300 projects were awaiting interconnection approval earlier this year, according to the complaint – the largest queue backlog of any regional transmission organization in the country.. Jacques Beaudry-Losique, founder of AI infrastructure and energy advisory firm GridSparq, said the broader power system is struggling to adapt to the speed and scale of AI-related demand.. “The AI-scale infrastructure build-out needs the quick deployment of a wide range of generation and transmission assets within reasonable time scales,” he told Data Center Knowledge. “We can’t do this very well today.”. Poorly Prepared for Next-Gen Data Centers. Beaudry-Losique argued that fragmented transmission ownership, interconnection delays, regulatory complexity, and years of underinvestment have left US electricity markets poorly prepared for hyperscale AI growth.. “I believe that there are maybe less than 1,000 people that fully grasp the mess we are in,” he said.. PJM has acknowledged the scale of the challenge in separate filings submitted to FERC earlier this year.. In a February filing tied to PJM’s proposed temporary auction price cap and floor revisions, the grid operator stated that “growing reliance on artificial intelligence (and associated data centers) has become a major driving factor for load growth.”. Related:Oklahoma Law Opens New Front in AI Data Center Power Debate. PJM also acknowledged that its previous market assumptions failed to anticipate the speed and scale of the current demand surge.. “The 2022 Quadrennial Review did not model the rapid and large increase in load growth … primarily associated with data centers,” PJM wrote.. The filing described current conditions as “unprecedented,” citing rapid load growth driven by high-demand data centers, along with generator retirements, queue congestion, permitting delays, and supply-chain constraints.. The market monitor is not arguing that data center growth alone created PJM’s current challenges. PJM’s own filings point to a combination of generator retirements, interconnection backlogs, permitting delays, supply-chain constraints, and compressed auction schedules. The dispute is whether AI-driven load growth merely exposed those weaknesses or fundamentally altered the market’s supply-demand balance.. Paul Patterson, an energy analyst at Glenrock Associates, cautioned against attributing PJM’s capacity market pressures solely to data center demand.. The analyst told Data Center Knowledge that the market monitor’s own report cites other factors, including capacity market pricing rules, uncertainty surrounding Effective Load Carrying Capability (ELCC) calculations, and lingering effects from previous interconnection queue management problems.. Those issues have become part of a broader debate over whether AI-driven load growth is the primary cause of PJM’s capacity crunch or whether it has exposed weaknesses already embedded in the market’s design.. Bring Your Own Power – A Vision of the Future?. The monitor argues that current market structures are poorly equipped to absorb concentrated, speculative load growth from hyperscale campuses.. Rather than continuing to integrate massive new data center demand directly into PJM’s existing capacity market structure, the report proposes a more radical approach: large data centers should effectively secure their own generation resources.. The report advocates for a “bring your own new generation” model, or BYONG, in which large loads secure generation resources through bilateral contracts or dedicated reliability mechanisms before interconnecting to the grid.. The proposal would shift more generation procurement risk onto developers. This represents a departure from traditional market structures in which new load is generally integrated into the broader grid.. “The current supply of capacity in PJM is not adequate to meet the demand from large data center loads and will not be adequate in the foreseeable future,” the report states. “There is not enough capacity currently to meet the data center load.”. Under that framework, future projects that fail to secure dedicated generation capacity could face curtailment during grid-stress events.. The monitor also warns against arrangements that would allow existing generation resources to bypass PJM markets and contract directly with data centers without broader reliability safeguards, arguing such deals could create “extreme price impacts” and reliability risks for other customers.. Travis Kavulla, vice president of regulatory affairs at NRG and former chairman of the Montana Public Service Commission, told Data Center Knowledge he broadly agreed with the market monitor’s concerns around speculative AI load forecasts and the growing need for dedicated-generation frameworks for hyperscale campuses, though he differed with parts of the report’s broader analysis and proposed remedies.. In a recent essay published in American Affairs, Kavulla argued that large AI campuses may increasingly need to be treated as a kind of “parallel industry” rather than conventional utility load, warning that traditional utility regulation and cost-allocation models were not designed for infrastructure deployments of this scale.. Beaudry-Losique said partially self-supplied campuses and dedicated-generation models are already emerging as developers pursue faster access to power capacity.. “Grid-connected, dedicated generation (a hybrid model) may very well be the market solution of the future,” he said.
