WASHINGTON: The Federal Energy Regulatory Commission on June 18, 2026 took its most aggressive step yet to clear the power bottleneck choking the AI buildout, issuing tailored “show cause” orders to all six US regional grid operators under Section 206 of the Federal Power Act. The orders give PJM Interconnection, the Midcontinent Independent System Operator, the Southwest Power Pool, the California Independent System Operator, ISO New England, and the New York Independent System Operator 60 days to either defend their existing interconnection frameworks or propose reforms that let large loads — specifically AI data centers — connect to the transmission system faster. TechCrunch called the action “a government-mandated fast lane to the grid.” The six operators also have 30 days to file informational reports showing that adequate generation is available for existing and new large loads in their regions.
The orders cover regions serving about 200 million Americans across more than 30 states and the District of Columbia, according to FERC. They do not apply to the fast-growing data center cluster in Texas, where ERCOT operates outside federal jurisdiction. The action stems from a rulemaking effort launched after Energy Secretary Chris Wright directed FERC in October 2025 to consider reforms aimed at speeding up large-load interconnection. Data centers themselves remain responsible for paying the costs of their own grid hookups under the new framework, a point regulators emphasized to address concerns that ratepayers would subsidize AI infrastructure.
Why the queue stopped moving
The numbers behind the orders explain the urgency. PJM, the largest grid operator in the country, received 811 new project applications representing 220 GW in a single April 2026 cluster study window — more than the entire installed generation capacity of most countries. More than 3,300 projects were already waiting in the PJM queue earlier this year. The median project takes more than three years to reach an interconnection service agreement and another four years to come online after approval, putting total wait times near a decade for some queue entrants.
The demand side is moving faster than the supply side can absorb. PJM disclosed it received 95 large-load adjustment requests totaling around 54 GW through November 2025, with utilities evaluating hyperscale AI campuses that need hundreds of megawatts each and, in some cases, gigawatt-scale single-site deployments. PJM’s own forecast shows the region adding 5 to 7 GW of data center demand annually from 2027 through 2032, against only 2 to 3 GW of new supply each year. That gap is the bottleneck FERC is now trying to force open.
What the orders actually require
The Section 206 mechanism is procedurally specific. By invoking it, FERC has put each operator in the position of having to affirmatively prove that its current tariff is “just and reasonable” without new large-load-specific provisions — a high bar given the queue data the agency already has. If an operator cannot defend the status quo, it must propose reforms that FERC will then evaluate. The 60-day clock starts running immediately.
The 30-day informational reports are the more immediate disclosure. Each operator must show what generation is actually available to serve existing and new large loads in its footprint. That filing will create the first apples-to-apples public dataset on the gap between data center demand and dispatchable supply across the six federal-jurisdiction grids. For founders and operators building infrastructure businesses on top of AI compute — colocation, edge deployments, hybrid cloud — those reports will be the single most useful planning document of the year.
What does the FERC order mean for founders building in AI?
The practical near-term effect for AI startups is location math. Companies have been quietly routing new compute commitments to ERCOT-served Texas precisely because the federal interconnection rules did not apply and the buildout there could move faster. The FERC orders narrow that arbitrage. If PJM, CAISO, and the others are forced to clear their queues, the cost and time penalty of building in the federal jurisdictions falls, and the case for concentrating data center load in a single state weakens.
The order also reshapes the calculation for power-side founders. The thesis behind the current wave of AI-adjacent energy raises — including recent funding rounds for advanced nuclear, geothermal, and grid-scale storage startups — is that dispatchable baseload is the binding constraint on the AI buildout. FERC just validated that thesis at the regulator level. Companies whose product is “more power, sooner” now have a federal mandate moving in their favor, which is a meaningful tailwind for raise conversations, customer contracts, and permitting timelines.
For everyone else, the watch item is rate impact. The orders explicitly require data centers to pay their own interconnection costs, but the broader question of how transmission upgrades get allocated between data centers and existing ratepayers remains open. Several commissioners and consumer advocates have warned that without strict cost-causation rules, AI infrastructure could end up subsidized by households and small businesses. How each operator handles that allocation in its 60-day response will set the political ceiling on how fast the buildout can actually move.
What’s next
The 30-day informational reports are due in mid-July 2026 and will be the first concrete data drop. The 60-day show-cause responses follow in mid-August. FERC will then evaluate each operator’s filing individually, with the option to accept the existing rules, accept proposed reforms, or impose its own framework if neither answer is adequate. Industry observers expect at least PJM and MISO to propose substantive reforms rather than defend the status quo, given the queue data already on the public record.
Three things to watch over the next 60 days. First, whether any operator attempts to push back on FERC’s jurisdictional reach — former FERC Chair Mark Christie previously called similar federal action “an unprecedented expansion of federal control” over state retail authority, and that argument may resurface in the show-cause responses. Second, whether Texas data center commitments slow as the federal-jurisdiction queue arbitrage narrows. Third, whether the AI hyperscalers — Microsoft, Google, Amazon, Meta, and the model labs — start publicly endorsing or filing in support of specific reform proposals, which would shift the political weight behind the fastest options. The outcome will set the pace of the entire AI compute buildout for the rest of the decade.
