The largest power grid operator in the United States just told the electricity industry something it has never said before: the system that keeps the lights on for 65 million people may not be able to keep up with artificial intelligence.
PJM Interconnection, which coordinates the flow of electricity across 13 states and the District of Columbia, announced in May 2026 that it is launching a sweeping redesign of its wholesale electricity markets to address what it calls unprecedented demand growth driven by data center construction and AI workloads. In its announcement, PJM used the phrase “years, not decades” to describe the window for action, though the operator has not attributed the line to a named individual or specific internal document.
That language matters. PJM is not a think tank issuing a white paper. It is the entity responsible for making sure there is enough electricity flowing through wires from New Jersey to Illinois every second of every day. When it says the current market design cannot attract enough new power plants to match what is coming, utilities, regulators, and the companies building AI infrastructure all have reason to pay attention.
What PJM is actually doing
PJM’s reform effort targets the core mechanisms that determine how power plants get paid and how new generation enters the market. The operator plans to convene stakeholders across its territory to reexamine three interlocking structures: capacity markets, energy markets, and ancillary services. Together, these are supposed to ensure reliability during peak demand. The goal, according to PJM’s announcement, is to sharpen pricing signals so they reflect real-time reliability needs and pull investment toward firm, dispatchable resources as well as storage and other flexible technologies.
PJM described this as a multi-year process but said some elements may be fast-tracked to address near-term risks.
At the federal level, the Federal Energy Regulatory Commission has already taken a related step. FERC accepted tariff revisions filed by PJM under docket ER26-1556. In a concurrence tied to that order, Commissioner Rosner stressed both the urgency of near-term fixes and the difficulty of longer-range structural changes. Rosner referenced PJM’s layered obligations to states, market participants, and federal regulators. The concurrence made clear that FERC’s acceptance of the tariff revisions should not be read as a final endorsement of PJM’s broader reform agenda. Future filings and contested proceedings are likely.
Together, these two actions form the strongest public evidence that institutional actors are treating AI-driven load growth as a live operational threat, not a planning abstraction.
Why data centers are different from normal load growth
Electric grids have always had to plan for growing demand. What makes the current moment unusual is the speed, scale, and geographic concentration of data center construction.
Northern Virginia alone hosts the largest cluster of data centers on the planet, and new campuses are spreading into central Ohio, suburban Chicago, and parts of Maryland. These facilities do not behave like typical commercial buildings. A single large AI training cluster can consume as much electricity as a small city, and it runs around the clock at near-full load. PJM’s public communications have begun treating data centers and AI clusters as a distinct planning category rather than a subset of generic commercial demand.
PJM’s interconnection queue, the pipeline through which new generators apply to connect to the grid, is another area where hard data remains limited in the public record tied to these latest actions. Grid operators across the country have struggled with backlogs that delay new projects by years. The market reform announcement and the FERC tariff order do not include detailed queue statistics or firm timelines for clearing the bottleneck, leaving open how quickly new supply can realistically come online.
What is still unclear
PJM has not released granular load-forecast models showing exactly how many megawatts of new AI and data center demand it expects over the next five to ten years. The “years, not decades” warning appears to be grounded in internal modeling that has not been made fully public. That does not make the warning less credible, but it does mean outside analysts cannot independently stress-test the assumptions behind it.
The resource mix question is also unresolved. PJM has not specified whether it expects the bulk of new capacity to come from natural gas plants, renewables paired with storage, nuclear facilities, demand response, or some combination. Each pathway carries different implications for construction timelines, fuel costs, emissions, and long-run price stability.
State-level permitting adds another layer of complexity. Even if PJM’s reformed markets send stronger financial signals, power plants and high-voltage transmission lines cannot be built without state and local approval. Several states within PJM’s footprint have their own renewable energy mandates, fossil fuel restrictions, or environmental review processes. Some may welcome data center-driven economic development; others may resist the land use, water consumption, or emissions that come with it. Commissioner Rosner’s concurrence acknowledged this tension without resolving it, leaving open whether FERC will approve the full scope of PJM’s proposals or push back where they conflict with state policy.
What this means for electricity bills and business decisions
For the 65 million people who depend on PJM’s grid, the near-term effects are indirect but real. Capacity auction prices have been rising, and further market redesign aimed at attracting new generation will likely push those costs higher. Those increases flow through to retail electricity bills via utility procurement and default service rates. Ratepayers in Virginia, where data center density is highest, may feel the pressure first, especially if local transmission constraints require expensive upgrades.
Businesses planning new data center investments in PJM’s territory face a different calculus. The operator’s willingness to overhaul its markets signals that it takes supply shortfall risk seriously, which could reassure companies worried about long-term power availability. But evolving market rules also mean that interconnection timelines, curtailment risks, and electricity costs could shift over the life of a project. Developers may increasingly need to budget for on-site backup generation or premium reliability services as a baseline requirement for mission-critical AI infrastructure.
Why PJM’s next moves will shape the national debate over AI power demand
PJM and FERC have put down clear markers: AI-driven demand growth is reshaping the planning landscape, and market rules will not stay the same. But the gap between policy intent and physical construction is where reliability risk actually lives. A reformed auction design can attract investment on paper. Whether new gas plants, battery storage facilities, or nuclear units break ground fast enough to match the load curve is a separate question, one that depends on permitting, supply chains, financing, and political will across more than a dozen states.
As of June 2026, PJM’s stakeholder process is underway, and the first concrete rule proposals are expected to surface in the coming months. Until those details emerge, the scale of the challenge and the distribution of costs remain only partially visible. What is no longer in doubt is that the operator responsible for the largest share of the American power grid believes the status quo will not hold.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
