By Shane Snider | Senior News Writer
July 31, 2026
Executive Overview
The operational landscape for artificial intelligence infrastructure developers in the United States is undergoing a profound structural shift. PJM Interconnection—the nation’s largest wholesale electricity market operator, managing the grid across 13 states and the District of Columbia—has formally approved a sweeping new regulatory framework designed to address the unprecedented surge in power demand driven by AI workloads and hyperscale data centers.
Approved on July 27 by the PJM Interconnection Board of Managers and slated for immediate filing with the Federal Energy Regulatory Commission (FERC), the policy marks a watershed moment in American energy policy. Under the proposed framework, future large-scale electricity consumers will no longer be able to assume that plugging into the regional grid guarantees unconditional, firm power service. Instead, new campuses must either proactively secure dedicated generation capacity or accept the legal risk of operational curtailments during periods of systemic energy scarcity.
The initiative represents one of the first aggressive attempts by a major U.S. grid operator to decouple physical grid interconnection from traditional resource adequacy. By forcing data center developers to prove their energy backing before enjoying firm-service economics, PJM is attempting to insulate existing ratepayers from soaring capacity costs while rewriting the playbook for how digital infrastructure is financed, built, and powered.
Detailed Chronology: How PJM Reached a Breaking Point
The convergence of artificial intelligence, high-performance computing, and legacy power grid architecture has been years in the making, but the crisis point within the PJM footprint has accelerated dramatically.
The Warning Signs and Market Shortfalls
The systemic stress on the PJM grid is not merely theoretical; it is reflected in recent market outcomes and capacity auctions:
- Generation Retirements vs. Load Growth: Since 2022, approximately 15 gigawatts (GW) of legacy fossil fuel and nuclear generation have retired across the PJM region, driven by environmental regulations and aging infrastructure. Simultaneously, PJM’s long-term forecasting models project roughly 70 GW of new large-load demand entering the system by 2038, overwhelmingly driven by hyperscale AI campuses.
- Underperforming Capacity Auctions: The disjunction between supply and demand became undeniable during recent capacity auctions. Two consecutive Base Residual Auctions cleared significantly below the system’s mandated reliability requirement. Most notably, the 2028–2029 Base Residual Auction produced a staggering 6,831-megawatt (MW) capacity shortfall, setting off alarm bells among utility regulators, federal officials, and grid operators.
- Federal Pressure and Warnings: The policy pivot follows intense external scrutiny. Earlier this year, the White House issued explicit warnings urging PJM to reform its grid governance structures "before it is too late," highlighting the systemic risk that unmanaged AI load growth poses to national economic stability and grid security. Furthermore, during severe regional heat waves, PJM was forced to issue unprecedented backup-generator warnings, exposing the thin margins separating baseline stability from rolling brownouts.
The Board’s Action and FERC Filing
Recognizing that business-as-usual planning would inevitably compromise system integrity, the PJM Board of Managers convened to draft a comprehensive corrective package. The resulting framework, approved on July 27, pivots away from passive load absorption toward active, commitment-first resource planning. PJM’s leadership will now submit the policy to FERC, where it is expected to face intense lobbying from both tech giants and traditional utilities.
Supporting Context & Metrics: The Mechanics of the Crisis
To understand why PJM is taking such drastic measures, one must examine the hard metrics defining the modern energy transition.
The Scale of the AI Power Demand
Hyperscale data centers are no longer comparable to traditional commercial real estate. While older enterprise data centers typically consumed between 10 to 30 megawatts, modern AI "factories" and training clusters frequently demand 100 to 500 megawatts or more per campus. Extrapolated across Virginia’s "Data Center Alley" (Northern Virginia), Ohio, Pennsylvania, and other hot spots within the PJM footprint, the cumulative load represents an industrial expansion unlike anything the grid has witnessed since the post-World War II manufacturing boom.
The Mechanics of IRAS (Interim Resource Adequacy Service)
At the center of PJM’s proposal is the Interim Resource Adequacy Service (IRAS), previously conceptualized under the moniker "Connect and Manage."
- Implementation Date: Set to take effect on June 1, 2027.
- Operational Rules: New large loads that fail to bring sufficient generating capacity to the system—or otherwise contract for bilateral supply—will be placed under the IRAS framework.
- Curtailment Priority: Unlike legacy customers who enjoy guaranteed firm service, IRAS-governed facilities will be legally designated as interruptible. During system resource shortages or grid stress events, these data centers will face mandatory curtailments before PJM triggers broader Pre-Emergency Load Management protocols.
- The Large Load Registry: To prevent shadow forecasting errors and aid state-level regulators, PJM is establishing a dedicated Large Load Registry. Because PJM lacks the jurisdictional authority to set retail electricity rates, this registry will provide state public utility commissions with the granular data needed to design appropriate retail rate classes and cost-allocation mechanisms.
The Reliability Backstop Procurement
To address the immediate 6,831-MW shortfall for the 2028–2029 delivery year, the PJM Board has ordered a one-time Reliability Backstop Procurement, scheduled to begin in September. This procurement mechanism is designed to acquire supplemental capacity outside the standard auction cycle.
Crucially, PJM has structured the backstop to be flexible: procurement targets will be dynamically reduced by signed bilateral contracts for new supply, approved new generation builds, and large-load sites that commit to formal demand-side participation. Developers participating in this process, however, will receive no preferential treatment and must navigate PJM’s standard, notoriously congested interconnection queue.
Official Statements & Industry Perspectives
The policy pivot has elicited strong reactions across the energy and technology sectors, highlighting a fundamental philosophical split between digital infrastructure developers and grid operators.
Redefining "Connected" vs. "Firm"
Neil Osnato, founder of Persistence Analytics Group, provided critical insight into the structural shift embodied by the PJM proposal. Speaking with Data Center Knowledge, Osnato emphasized that the policy permanently severs physical grid connectivity from financial resource adequacy.
"PJM is beginning to separate ‘connected’ from ‘firm,’ and asking new large loads to earn the reliability product they previously could largely assume," Osnato explained.
He cautioned technology companies against expecting favorable regulatory treatment without physical proof of performance:
"Do not grant firm-service economics against an unverified capacity or load claim. Prove the megawatts, prove the flexibility, then grant the firmness."
According to Osnato, power is no longer merely a utility service purchased at the wall; it has become a core component of the hyperscale capital strategy. Developers will increasingly be forced to invest directly in dedicated generation, battery energy storage systems (BESS), virtual power plants (VPPs), and microgrids to ensure operational continuity.
Preventing Cost-Shifting to Ordinary Ratepayers
In its official filings and board commentary, PJM emphasized that equity and consumer protection served as primary motivations for the new rules. The Board explicitly warned that existing residential and commercial ratepayers must not be forced to subsidize the infrastructure required to power trillion-dollar AI models.
"Existing consumers should not bear higher capacity costs caused by new large loads that do not bring, or otherwise contract for, the new supply necessary to serve them," the Board noted. While PJM acknowledges it lacks the direct authority to dictate retail rate structures, the framework empowers state regulators to shield everyday consumers from surging utility bills.
A National Blueprint?
The ramifications of PJM’s proposal extend far beyond its 13-state footprint. Rob Gramlich, president of Grid Strategies, noted that other regional transmission organizations (RTOs) are closely watching PJM’s experiment as a potential template for their own impending crises.
"Texas (ERCOT), MISO, and SPP are considering similar policies," Gramlich told Data Center Knowledge. As power constraints pop up across North America, the concept of "commitment-first planning"—where load growth must be preceded by ironclad supply commitments—is rapidly becoming the new industry standard.
Future Outlook: What Lies Ahead for AI Infrastructure
As FERC evaluates PJM’s proposal, data center developers, utility executives, and investors must adapt to a dramatically altered operational paradigm. The era of speculative site selection—where developers secured grid interconnection queues years in advance with little more than a napkin sketch of their power requirements—is drawing to a close.
1. The Rise of Behind-the-Meter and Bilateral Solutions
To bypass the regulatory bottlenecks and curtailment risks of the IRAS framework, hyperscalers are expected to double down on direct procurement. This includes striking direct bilateral power purchase agreements (PPAs) with nuclear and renewable energy operators, co-locating data centers directly next to power generation facilities (such as behind-the-meter nuclear deals), and investing heavily in advanced energy storage.
2. Increased Project Costs and Timelines
Capital expenditures for AI infrastructure will inevitably rise. Developers will need to factor the cost of securing or building dedicated capacity directly into their real estate models. Consequently, site selection will increasingly favor regions where surplus generation already exists, potentially shifting data center development away from saturated markets like Northern Virginia toward regions with robust local power supplies.
3. Regulatory and Legal Battles at FERC
The upcoming FERC review will be a high-stakes legal battle. Tech sector lobbyists are expected to challenge aspects of the IRAS framework, arguing that penalizing new loads stifles American technological competitiveness in the global AI race. Conversely, consumer advocacy groups and traditional utilities will push for even stricter compliance standards to protect grid reliability.
Ultimately, PJM’s framework serves as an unmistakable warning to the digital economy: the physical laws of the electrical grid cannot be bypassed by software optimization. For the AI boom to continue uninterrupted, the tech sector must become an active builder of the power grid, rather than a passive consumer of its dwindling reserves.
