AI data centers Put FERC Grid Rules Under Strain

AI data centers near transmission towers at dusk

AI data centers have moved from a site-selection issue to a transmission-planning problem for federal regulators. On June 18, 2026, the Federal Energy Regulatory Commission issued show cause orders under Section 206 of the Federal Power Act to all six U.S. Regional Transmission Organizations and Independent System Operators, directing them to justify or reform rules for how large energy users connect to the transmission grid, according to FERC’s June 18 notice.

The action did not create one national data center interconnection tariff on that date. It did, however, put PJM, MISO, SPP, CAISO, ISO-NE, and NYISO on a formal path to explain whether their existing rules can handle large loads whose size, timing, and power behavior differ from many historical industrial users. Each region was directed to submit reports within 30 days detailing how adequate generation would be secured for existing and new large loads.

The issue is partly technical and partly financial. Large computing facilities can require transmission service, substation work, generation planning, and operational studies before racks are installed. If the project is delayed, resized, or withdrawn, the grid operator may already have spent time and money studying capacity that never appears. Prior Abacus coverage of data center energy examined how these questions have shifted from local utility planning into federal tariff design.

Why AI data centers Complicate Interconnection

AI data centers And Faster Load Ramps

FERC’s regulatory attention followed a change in project scale. Research notes cite FERC market material showing that average new data centers in service in 2025 were nearly 80 MW, compared with around 25 MW in 2020. That change matters because interconnection studies are not only about annual electricity consumption. They also account for local deliverability, system peaks, generation availability, and whether transmission assets can support demand at specific nodes.

For AI data centers, the operational concern is not limited to total megawatt demand. The research record points to fast ramp-up in power demand and high urgency for connection as traits that differ from traditional loads. Those traits can reduce the usefulness of planning models that assume slower industrial expansion or more predictable construction timelines. A project that requests a large block of capacity before its computing deployment is fully certain can create planning noise across the regional queue.

This does not mean every computing site creates the same grid burden. The available research does not provide site-level load shapes, power usage effectiveness figures, or cooling designs. It does show that FERC treated the category of significant electrical loads as large enough to require regional tariff scrutiny. That is a narrower claim than saying data centers alone caused all interconnection pressure.

Queue Discipline Becomes A Planning Issue

Interconnection queues were already large before the June 2026 orders. The 2024 FERC State of the Markets report, released in March 2025, put active interconnection requests in RTO and ISO queues at about 2,289 GW at the end of 2024, down from roughly 2,368 GW in 2023. Those figures show a system where study resources and planning assumptions were already stretched.

AI data centers also create a different kind of queue concern: speculative load. FERC’s new orders included mechanisms for escalating readiness requirements and withdrawal penalties aimed at reducing requests that may never turn into operating facilities. That approach is not a ban on ambitious projects. It is an attempt to make sure interconnection positions reflect a credible ability and intent to connect, not only a desire to reserve scarce grid capacity.

What FERC Ordered On June 18, 2026

Show Cause Orders Across Six Grid Regions

The June 18 orders used FERC’s Section 206 authority, which allows the commission to examine whether existing jurisdictional rates, terms, or practices are just and reasonable. In this case, the targeted rules govern how large loads connect to the transmission grid in organized markets. The regions named in the research record were PJM, MISO, SPP, CAISO, ISO-NE, and NYISO.

That regional structure matters. The United States does not have one operational grid market. RTOs and ISOs use different tariff structures, queue practices, generation accreditation rules, and regional planning methods. FERC’s move therefore sought responses from each operator rather than imposing a single technical interconnection template at the outset.

The 30-day report requirement also put generation adequacy near the center of the proceeding. Data center interconnection is not only a wires question. If new load appears faster than generation or transmission upgrades can support it, reliability and cost allocation become linked. FERC’s demand for information on how adequate generation will be secured reflects that connection.

Cost Recovery Agreements Shift Financial Risk

One of the most consequential features in the research record was the use of Cost Recovery Agreements. These agreements are designed to make large loads financially responsible even if they fail to come online, reducing the risk that study, upgrade, or preparation costs migrate to residential ratepayers.

If AI data centers reserve capacity and then withdraw, the cost issue can be significant even without a completed facility. Grid operators may have processed studies, evaluated transmission impacts, or planned upgrades around the requested load. Cost recovery tools are intended to place more of that risk on the party seeking service, rather than spreading it across customers who did not cause the request.

The technical effect is indirect but material. Stronger financial readiness screens can reduce queue congestion, improve the signal value of load forecasts, and help planners distinguish advanced projects from early-stage proposals. The research does not show whether the June 2026 framework will cut queue times in practice. That will depend on how each region designs its tariff response and how FERC evaluates those filings.

Regional Tariffs Are Becoming The Control Point

Transmission lines crossing an industrial development site

PJM And Co-Located Generation

FERC had already acted before the June 2026 show cause orders. In December 2025, it directed PJM to create clearer tariff rules for large loads co-located with generation. The decision came after concerns that existing rules lacked transparency and consistency and did not allocate cost responsibilities clearly enough. PJM is especially significant because it serves more than 67 million people across 13 states and the District of Columbia.

Co-location can appear simple from a project perspective: place load near generation and reduce dependence on distant transmission. In tariff practice, the questions are harder. Grid operators still need to know whether power is behind-the-meter, delivered through the transmission system, reserved during outages, or backed by network service. The research notes do not provide PJM’s final tariff design, so the reliable conclusion is narrower: FERC required clearer rules because the existing structure was judged insufficient for emerging large-load arrangements.

SPP’s High Impact Large Load Process

The Southwest Power Pool provides another example of regional experimentation. In January 2026, FERC approved SPP’s High Impact Large Load initiative, known as HILL. The initiative established an expedited study process for large loads and associated generation that are electrically proximate.

That design addresses one recurring problem: large load and generation may be proposed together, but conventional processes can study them through separate tracks or mismatched timelines. An expedited process for electrically proximate resources may reduce some study friction. The research record does not establish that HILL will solve broader transmission constraints, and it does not show whether similar models will be adopted by every region.

Readers interested in further technology and policy developments can also explore related content at Way Latino, a site within the same network.

FERC’s AI Data Center Interconnection Test

What The Orders Do Not Resolve

The June 2026 action was significant, but its limits are as important as its scope. Based on the research record, FERC did not set server efficiency standards, prescribe data center electrical architecture, or decide where individual facilities should be built. Its intervention focused on transmission tariff rules, cost responsibility, readiness requirements, queue discipline, and the link between new large loads and generation adequacy.

That leaves several unresolved technical questions for utilities, developers, state regulators, and regional operators. They still need to assess substation capacity, local transmission constraints, backup arrangements, cooling-related power demand, and the timing of generation additions. Federal tariff reform can improve process discipline, but it cannot remove the physical requirement to deliver power reliably at a specific location.

Who Bears The Operational Burden

The affected parties are broader than data center developers. Residential customers face exposure if tariff rules allow abandoned or downsized projects to shift costs into general rates. Grid operators face modeling burdens when large requests enter queues before firm construction plans are visible. Generators and transmission owners need clearer signals about whether new demand is durable enough to justify investment.

For AI data centers, the practical message is that interconnection has become part of infrastructure strategy, not a late-stage utility formality. FERC’s role is moving toward enforcing clearer regional rules before large loads distort planning assumptions or cost allocation. The evidence available as of October 18, 2026 supports a cautious reading: federal action has tightened scrutiny, but the operational results will depend on the tariff filings, readiness tests, and cost recovery terms that each regional market applies.

Related Articles