PJM Interconnection is reviewing its market and planning rules after nearly 4 gigawatts of anticipated data center load – roughly the output of four large nuclear reactors – failed to materialize on its system, a development that undermines the load-growth narrative driving capacity prices, transmission build-out, and generation retirement decisions across the Mid-Atlantic and Midwest.
The missing gigawatts and why PJM is reacting
The 4 GW figure represents load that had been included in PJM’s planning models – likely through the Large Load interconnection process or capacity market commitments – but which never actually connected or consumed power. In PJM’s 2024 load forecast, data centers accounted for the vast majority of projected demand growth, with some scenarios showing 15-20 GW of new data center load by 2030. A 4 GW disappearance in a single episode is not a rounding error; it is roughly 2.7% of PJM’s 2023 peak demand of 147 GW and equals the entire peak demand of a state like Delaware or New Hampshire.
PJM’s “Large Load” process, reformed in 2023, was designed to give hyperscale customers a faster path to interconnection while protecting other market participants from stranded costs if those loads don’t materialize. The process requires milestone payments, study deposits, and commercial operation date commitments. That such a large block of load could still exit – an “Irish goodbye” in the source’s phrasing – suggests either the financial commitments were insufficient, the contractual mechanisms have gaps, or the load was never as firm as the queue position implied.
The review PJM has launched will likely examine whether milestone schedules, financial security requirements, and cost-allocation rules need tightening. It may also revisit how speculative load is treated in the Regional Transmission Expansion Plan (RTEP) and the capacity market’s Reliability Pricing Model (RPM), where forecasted load directly sets the demand curve and thus clearing prices. If 4 GW of assumed load vanishes, the demand curve shifts left, potentially suppressing capacity prices by hundreds of millions of dollars across a delivery year – money that would have gone to existing generators and new capacity resources.
What this reveals about the data-center-demand thesis
The dominant industry narrative since 2022 has been that hyperscale data centers – driven by AI training, cloud expansion, and digital infrastructure – will deliver structurally higher electricity demand for a decade or more. Utilities from Dominion to AEP to FirstEnergy have cited this growth to justify keeping coal plants online, building new gas peakers, and accelerating transmission investment. PJM’s own 2024 Load Forecast Report showed summer peak demand growing 1.4% annually through 2039, with data centers as the primary driver.
That thesis rests on two assumptions: that hyperscalers’ announced projects translate into firm, metered load on a predictable timeline, and that the load, once connected, is relatively inelastic and long-lived. The 4 GW exit challenges both. Hyperscalers routinely secure optionality – land, interconnection queue positions, power purchase agreements – across multiple regions simultaneously, committing capital only when specific campuses reach final investment decision. A single company’s portfolio decision can swing multiple gigawatts. Microsoft, Google, Amazon, and Meta each have multi-gigawatt pipelines; a strategic pause by one can look like a systemic forecast error.
By comparison, the ERCOT region has seen similar dynamics: the “large flexible load” category in its interconnection queue exceeds 100 GW, but only a fraction reaches commercial operation. CAISO’s queue tells a comparable story. The difference in PJM is the scale of the planning apparatus – capacity market, RTEP, state resource adequacy mandates – that treats queue entries as de facto commitments. When they prove not to be, the cost spreads across all load-serving entities.
There is also a contract-structure dimension. Many hyperscalers prefer “sleeved” PPAs or virtual PPAs that settle financially but don’t require physical delivery to a specific PJM node. Others negotiate special tariffs or direct retail arrangements that bypass the capacity market entirely. If 4 GW of load was counted in PJM’s models but structured to avoid capacity obligations, the market has been over-procuring reliability resources. That points to a potential misalignment between how load is modeled for planning and how it is contracted for commerce.
Who this affects
- Utility resource planners: Integrated resource plans filed in 2024-2025 across PJM states likely embed the 4 GW (and more) as firm load. Planners should stress-test portfolios against a “low data center realization” scenario and document the capacity price exposure if demand curves shift.
- Generation developers (gas, solar, storage, wind): Capacity market revenues underpin project finance for new resources. A sustained reduction in the RPM demand curve could lower clearing prices by $20-$50/MW-day in affected delivery years, altering pro forma returns for projects targeting 2026-2028 commercial operation.
- PJM market monitor and FERC: The episode may trigger a Section 206 investigation or stakeholder process to reform Large Load milestone requirements, financial security, and cost allocation for network upgrades triggered by speculative load.
- Hyperscale customers and their energy advisors: Expect tighter contractual terms: higher milestone deposits, firmer commercial operation date penalties, and possible “use-it-or-lose-it” capacity reservation fees. The era of low-cost queue optionality in PJM is ending.
- Transmission owners and state regulators: RTEP projects justified by data center load may face prudence challenges if the load doesn’t materialize. State commissions should require utilities to demonstrate load realization before approving rate recovery for transmission upgrades driven by specific large-load assumptions.
What to watch next
- PJM stakeholder process timeline and scope: The Markets & Reliability Committee and Planning Committee will likely open a dedicated workstream. Track whether the review addresses only Large Load process rules or also revisits RPM demand curve construction and RTEP load assumptions.
- Capacity auction results for 2025/26 and 2026/27 delivery years: If the 4 GW removal is reflected in the 2025 Load Forecast Report (due early 2025), the Base Residual Auction demand curve will shift. Watch for clearing price deviations from the 2024/25 BRA ($28.92/MW-day for RTO) that exceed what fuel prices alone would explain.
- Hyperscale earnings calls and capital expenditure guidance (Q4 2024 – Q1 2025): Microsoft, Alphabet, Amazon, and Meta will signal whether the slowdown is PJM-specific or portfolio-wide. A broad capex revision would imply the 4 GW is the leading edge of a larger forecast correction.
- FERC docket activity on large-load cost allocation: Any complaint filing by existing load-serving entities or generators alleging unjust cost shifts from unmaterialized load will signal regulatory appetite for retroactive remedies.
- State-level IRP updates in Virginia, Ohio, Pennsylvania, New Jersey: These states host the bulk of PJM data center growth. Regulators may reopen or supplement recent proceedings to interrogate load assumptions.
Bottom line
The 4 GW exit is not a one-off anomaly; it is the first hard evidence that the data center demand curve embedded in PJM’s planning and market constructs is materially softer than modeled. Until Large Load rules align financial commitments with planning consequences, every gigawatt in the queue should be treated as probabilistic, not promised – and capacity prices, transmission plans, and resource adequacy assessments should reflect that uncertainty explicitly.
Read the full report at Renewable Energy World
Note: facts and figures attributed above to Renewable reflect that outlet's original reporting. Broader context, cross-sector connections, and forward-looking scenarios reflect independent analysis by our editorial team.
About this article: Drafted by Energy Ai with AI-assisted research and writing based on public reporting, then reviewed under our editorial process before publication.
Leave a Reply