CRITICAL EVENT UPDATE · AI Infrastructure × Power Grid × Regulation
FERC Approves PJM’s Data-Center Registry but Rejects Priority Curtailment, Adding Transparency Without Solving the Power Gap
Critical Event Update | Large-Load Registry × Grid Reliability × AI Infrastructure
The Federal Energy Regulatory Commission accepted PJM’s large-load registry but rejected the core plan to curtail new data centers lacking dedicated power before other large users. The decision creates a more visible demand database without resolving PJM’s roughly 6.8 GW 2028/29 capacity shortfall, its forecast of about 29 GW of data-center load by 2030, or cost allocation.
PJM forecast for data-center load by 2030
PJM's 2028/29 capacity shortfall
PJM large-load definition
Increase in PJM capacity costs since 2024, per Reuters
Registry approved | Priority-curtailment proposal rejected | Roughly 6.8 GW shortfall remains | About 29 GW of data-center load forecast by 2030
01 · RESEARCH BRIEF
The one-minute brief
In Docket ER26-3515, FERC accepted PJM tariff revisions for a large-load registry and rejected parts of the proposed curtailment structure that lacked sufficient jurisdictional and non-discrimination support. The registry will capture provider, in-service date, ramp schedule, backup generation, fuel and location. For ACIS, the decision strengthens the view that Time-to-Power is the first commercialization gate for AI infrastructure: visibility improves, but physical capacity, interconnection, cost allocation and reliability governance remain unresolved.
Audio transcript
FERC approved PJM's large-load registry but rejected the first priority-curtailment plan for data centers. Transparency improves, yet the roughly six-point-eight-gigawatt capacity shortfall and the forecast of about twenty-nine gigawatts of data-center load by 2030 remain. Time-to-Power, executable power contracts and cost allocation are still the key constraints.
Known facts and open questions
- Confirmed
- FERC accepted PJM's large-load registry effective 12 October
- Confirmed
- FERC rejected the core priority-curtailment terms while allowing a revised filing
- Confirmed
- The registry captures location, in-service timing, ramp, backup generation and fuel
- Unresolved
- Physical capacity, interconnection timing, cost allocation and final curtailment rules
Transparency
A unified registry improves project, location, in-service and ramp data
Reliability
A roughly 6.8 GW shortfall remains; a registry does not create power
Regulation
FERC rejected priority curtailment pending stronger jurisdictional and discrimination support
AI Infrastructure
Time-to-Power and executable power contracts still precede GPUs in revenue conversion
Capital
Delayed energization extends interest carry, depresses utilization and defers cash flow
02 · FACTS → IMPACT → VIEW
Why does this change matter?
Grid transparency improves, but the physical gate on AI infrastructure remains
- What is confirmed
- PJM had already established that fast large-load growth and resource retirements were pressuring reliability and affordability.
- Why it matters
- FERC approved the registry but rejected the first priority-curtailment structure. PJM still forecasts about 29 GW of data-center load by 2030 and reports a roughly 6.8 GW 2028/29 shortfall.
- ACIS view
- ACIS raises Grid Data Transparency while keeping Power Availability and Time-to-Power as hard constraints. Bring Your Own Power is a policy direction, not a settled federal implementation.
03 · EVIDENCE & ANALYSIS
Evidence and analysis
01｜What Happened
FERC accepted PJM's registry for large loads of at least 50 MW, including provider, in-service date, ramp schedule, location, backup generation and fuel. It rejected the core terms that would have curtailed new data centers without dedicated supply ahead of other large users, finding the distinction insufficiently supported. PJM may submit a revised proposal.
02｜Why It Matters Now
PJM forecasts about 29 GW of data-center load by 2030, roughly triple today's level. It reports an approximately 6.8 GW shortfall for the 2028/29 capacity year, while capacity costs have increased more than 1,000% since 2024. Demand visibility is now a reliability, price and financing issue.
03｜Confirmed Facts vs Uncertainty
Confirmed: the registry was accepted, priority curtailment was rejected and PJM may revise. Uncertain: the final curtailment order, state-versus-federal cost allocation, qualification of dedicated supply, duplicate-project volume and whether registry data accelerates interconnection.
04｜Transmission Mechanism
Better registry data → better load forecasts → more grounded resource and transmission planning; unresolved shortfall → elevated capacity, long-term power and backup-supply costs → delayed data-center commissioning → higher interest carry and lower utilization → slower AI-compute supply and enterprise-AI deployment.
05｜Prior ACIS View → New Evidence → Updated View
Prior: the AI-infrastructure constraint was shifting from GPUs toward power, permits and financing. New evidence: FERC created the data layer but did not approve PJM's first priority-curtailment framework. Update: raise Grid Data Transparency; keep Power Availability, Cost Allocation and Time-to-Power unchanged, with greater regulatory execution risk.
06｜Cross-Asset / Cross-Industry Read-through
Grid equipment, dedicated generation, storage and powered campuses gain relative value. Data-center developers without executable power paths, highly levered projects and long-dated compute contracts face pressure. Utilities, industrial users and households remain exposed to cost-allocation disputes; credit markets should distinguish energized assets from planned megawatts.
07｜What Does NOT Change
Registry inclusion is not interconnection approval and creates no generation, transmission or capacity. FERC's rejection does not confer unconditional service on data centers or remove state permits, utility connection, grid-upgrade or dedicated-supply requirements.
08｜Risks / Alternative Scenarios
Upside: the registry removes duplicate projects and resource additions arrive faster. Base: PJM refiles a narrower reliability proposal while connection terms tighten. Downside: the shortfall grows, states act first and projects stall or shoulder more generation and transmission costs.
09｜Next Validation
24H: final FERC order and PJM implementation notice. 7D: registry fields, covered projects and timeline for a revised curtailment proposal. 30D: capacity-auction signals, queue cleanup, state cost allocation and project-level power contracts and in-service dates.
10｜Research Classification
Critical Event Update + Research Memo + Decision Escalation. The event changes grid governance and evidence quality, not immediately available power.
11｜Human Decision Required
No automated asset action. Research governance should separate registered load, contracted power, interconnection approval and energized capacity in AI Infrastructure Weekly and Data Center Unit Economics.
04 · INVESTMENT IMPLICATIONS
Industry and asset implications
AI data centers
Value depends more heavily on an executable power path than planned megawatts.
Utilities and grid equipment
Planning visibility improves while capex and cost-allocation pressure rises.
Credit and financing
Energization delays widen interest carry and cash-flow mismatch.
Enterprise AI
Compute delivery remains constrained by regional power bottlenecks.
Transparency improves; supply, cost and timing constraints remain adverse.
05 · VALIDATION & RISKS
What to watch next
24H
PJM confirms implementation scope and registry mechanics
What would weaken the view: The final order or implementation differs materially from public reporting
7D
PJM publishes a compliant revision path and detailed fields
What would weaken the view: Rule uncertainty or state conflict expands
30D
Duplicate projects fall and resource or interconnection timing improves
What would weaken the view: The shortfall, costs or project delays continue to widen
What would change our view?
Treating registry approval as new supply would overstate AI-project delivery. Treating rejection of the first curtailment plan as unconditional data-center service would understate PJM revisions, state action and cost-allocation risk.
06 · FAQ
Key questions
Did FERC approve priority curtailment of data centers?
No. FERC approved the large-load registry but rejected the core priority-curtailment terms in the first proposal.
Will the registry solve PJM's power shortage?
No. It improves demand data and planning but adds no generation, transmission or capacity.
How much data-center load does PJM expect?
Reuters cites PJM's forecast of about 29 GW by 2030, roughly triple today's level.
What does this mean for AI infrastructure?
Time-to-Power, long-term power contracts, dedicated supply and interconnection certainty become even more important.
07 · TERMS & SOURCES
Terms, sources and related research
Key terms
- Large Load Registry
- A data mechanism recording the location, timing, ramp and supply arrangements of large electricity users.
- Curtailment
- Reducing or interrupting load under reliability or supply stress.
- Time-to-Power
- The time required for a project to obtain durable electricity service.
[2] Reuters｜U.S. panel approves PJM registry for data centers, rejects curtailment plan｜2026.10.12 ↗
This report relies on the public FERC order, PJM materials and Reuters reporting. The regulatory decision and PJM forecasts are confirmed; final curtailment rules, state cost allocation, project interconnections and actual energized capacity remain uncertain. 29 GW is not energized capacity, and 6.8 GW refers to a specific capacity year.