FERC Orders Grid Operators to Fast-Track AI Data Center Power Connections or Explain Why They Won't

For most of its history, the Federal Energy Regulatory Commission has operated in relative obscurity — a technocratic agency managing the plumbing of the American power grid, adjusting transmission tariffs, reviewing generator interconnection queues, and mediating disputes between utilities. The AI buildout has changed that. FERC has emerged as one of the most consequential federal regulators for the technology industry, and its latest action underscores exactly why.

The commission has issued show-cause orders to all six regional grid operators under its jurisdiction, requiring each one to either justify why its current tariffs and interconnection rules are adequate for large electricity users — specifically, AI data centers seeking “hundreds of megawatts, and in some cases gigawatt-scale, power commitments” — or propose changes within 60 days. A separate 30-day deadline applies to capacity planning: operators must also report on whether adequate generation capacity exists to serve these loads without compromising grid reliability.

FERC Chairman Laura V. Swett framed the action in terms of competing national priorities: “We need speed to power for large users while protecting consumers from bearing inappropriate costs.” It is a tension that has no easy resolution, and the commission’s orders represent a recognition that the current framework for connecting large loads to the grid was not designed for this moment.

The Six Operators and What They Control

The show-cause orders cover the six independent system operators and regional transmission organizations that manage the US grid outside Texas:

PJM Interconnection is the largest, covering 13 states from Illinois to New Jersey and serving roughly 65 million people. PJM operates the grid that powers the mid-Atlantic corridor, which has emerged as one of the most congested and contested regions for data center development due to its proximity to major internet exchange points and population centers.

Midcontinent Independent System Operator (MISO) covers a swath of states from North Dakota through Louisiana, an area with abundant wind resources and growing interest in AI campuses seeking access to cheap renewable power.

Southwest Power Pool (SPP) covers 17 states across the central and western US. SPP has already moved ahead of FERC’s order in some respects, implementing a 90-day expedited review process for large loads that agree to bring their own generation — a structure that FERC’s order may expand to other regions.

California ISO (CAISO) operates the grid for most of California, where the state’s aggressive renewable buildout and persistent transmission constraints create a particularly complex environment for large new loads.

ISO New England covers the six New England states, a region with aging infrastructure and historically high wholesale electricity prices that have made it a difficult market for data center siting.

New York ISO (NYISO) manages New York’s grid, subject to some of the most complex electricity markets and environmental regulations in the country.

Texas’s ERCOT, the country’s largest grid by generation capacity, is largely outside FERC’s jurisdiction due to its status as an intrastate system.

Why the Queue Has Become Ungovernable

The immediate trigger for FERC’s intervention is the state of interconnection queues — the waiting lists that generators and large loads must join before they can connect to the grid. Those queues have become effectively non-functional. At the end of 2025, there were more than 2,700 gigawatts of capacity in interconnection queues across the US — more than twice the total installed generating capacity of the entire country. Most of this capacity will never be built; the queues are clogged with speculative projects that hold their place for years while preventing viable projects from moving forward.

AI data centers seeking grid connections have encountered this problem acutely. A company planning a 500-megawatt data campus — a moderate size by current standards, well below the gigawatt-scale facilities being proposed by Microsoft, Amazon, and Google — may be told it faces a multi-year wait for an interconnection study, followed by additional years for the physical connection to be built and energized. By the time the connection is ready, the computing technology the campus was designed to house may already be obsolete.

FERC’s order targets five specific areas each operator must address: faster application processes, cost-shifting safeguards to protect retail customers, co-location and behind-the-meter generation rules, flexible load transmission services for customers willing to operate with some power curtailment, and generation study processes.

The Consumer Protection Tension

The chairman’s explicit reference to protecting consumers from “inappropriate costs” points to a real political tension embedded in these proceedings. When large data centers connect to the grid, they often require transmission upgrades and new infrastructure that the interconnection rules typically require the connecting party to pay for. But the definitions of what must be paid by the connecting customer versus what gets socialized across all ratepayers are frequently contested — and the stakes are enormous.

A gigawatt-scale data center requiring a new 500-kilovolt transmission line and several substation upgrades could trigger costs of hundreds of millions of dollars. If those costs are allocated to ratepayers rather than the data center operator, they show up in the electricity bills of ordinary households and small businesses. As data center demand grows — and the power requirements of AI training clusters are expected to continue growing for years — this cost-shifting question will become increasingly politically charged.

Several states have already passed or are considering legislation requiring large electricity customers, specifically data centers, to demonstrate that their interconnection will not raise rates for existing customers. FERC’s order will need to navigate this tension at the federal level, setting rules that apply across regions with very different existing rate structures and regulatory environments.

The National Security Overlay

FERC’s action comes in the context of broader federal prioritization of AI infrastructure. The White House executive order on AI innovation and security, issued earlier this month, explicitly identified reliable power supply as a constraint on American AI competitiveness and directed federal agencies to coordinate on accelerating energy infrastructure for AI campuses.

The logic is straightforward: if American hyperscalers cannot secure power at the speed required to stay ahead of Chinese AI infrastructure buildout, the US competitive advantage in frontier AI erodes. The connection between megawatts and geopolitical position has been made explicit at the highest levels of the administration, transforming what was previously a dry regulatory question about grid interconnection into something closer to a national security imperative.

FERC is not the only agency acting. The Department of Energy has directed the national laboratories to develop methodologies for faster environmental review of transmission projects. The Army Corps of Engineers has established expedited review processes for transmission corridor permits on federal lands. The overall direction is unmistakable: the federal government has decided that AI infrastructure cannot wait for the normal pace of utility regulation.

What Comes Next

The 60-day window for grid operators to respond will close in mid-August. FERC is expected to finalize rules — incorporating or rejecting the operators’ proposed changes — by the end of 2026. The process will involve extensive comment periods from utilities, large electricity customers, consumer advocates, and environmental groups, each with competing interests in how the rules are drawn.

The practical effects, if the rules are changed as FERC seems to intend, would accelerate the timeline for connecting large AI facilities to the grid from the current multi-year horizon to something closer to twelve to eighteen months for projects that have their financing, permits, and generation sources in hand. That is still not fast by the standards of a technology industry accustomed to deploying software globally in days, but it represents a material improvement in an infrastructure bottleneck that has been constraining American AI expansion.

The harder question — who pays, and for what — will outlast the current proceedings. But the commission has made clear that the status quo of multi-year queues and undefined cost responsibilities is no longer acceptable. In the age of AI, the grid has become infrastructure that matters too much to be governed by rules written for a different era.