FERC Orders Six Grid Operators to Fix the Data-Center Queue

The shortage that limits how much computing the hosting and cloud industry can sell is no longer chips. It is the wait to connect a large new load to the electric grid, a queue measured in years, and on June 18 the US Federal Energy Regulatory Commission took the most consequential action yet to shorten it. FERC issued six tailored show-cause orders, one to each of the regional grid operators under its jurisdiction, under Section 206 of the Federal Power Act, giving them 60 days to either justify why their current rules for connecting large loads remain just and reasonable or file the changes needed to fix them. The targets are the operators that run the grid for most of the country: PJM, the Midcontinent ISO, the Southwest Power Pool, the California ISO, ISO New England, and the New York ISO.

The framing from the Commission was unusually direct. Commissioner David LaCerte told the operators, “I say this not as a threat, but as a statement of duty.” For anyone whose business depends on new data-center capacity coming online on schedule, this is the regulatory event of the quarter, and the cost-allocation fight buried inside it will set the floor price of compute capacity for years.

Key facts, FERC large-load interconnection action (June 18, 2026)

• Six Section 206 show-cause orders, one each to PJM, MISO, SPP, CAISO, ISO-NE, and NYISO, directing each to defend its current large-load interconnection tariff as just and reasonable or file reforms
• The clock: resource-adequacy status information due within 30 days; substantive response or tariff filing due within 60 days; operators may request an extension, putting the outer bound at several months from the order date
• Five required reform areas: faster transmission-service application and study processes including alternative transmission technologies; prevention of cost shifts onto other ratepayers, with transparency on transmission costs; rules for co-location and behind-the-meter generation; new transmission services for flexible large loads; and a study process for generation serving “electrically proximate” co-located loads
• The scope gap: FERC only ordered the organized markets it regulates; it encouraged, but could not compel, transmission owners in non-RTO regions to act, and roughly one-third of Americans live in those regions
• The context: the orders are FERC’s answer to the Department of Energy’s earlier advance notice of proposed rulemaking on large-load interconnection, and arrive amid record data-center power demand and multi-year connection queues

Why the Queue, Not the GPU, Is the Constraint

For two years the public story of AI infrastructure has been a chip story: who can get Nvidia’s latest, how many, how fast. That constraint has eased relative to the one that replaced it. The binding limit now is power, and specifically the time it takes to get a new large load connected to the transmission system, which in the worst-affected regions has stretched to years. Interconnection rules are the gatekeeper between a planned data center and operating capacity, and those rules were written for a slower era, when large new loads were rare and a multi-year study process imposed no real cost. When the load is a gigawatt-scale AI campus and the queue is the constraint on national compute supply, the same deliberate process becomes the bottleneck, and FERC has now told the six operators that run it to defend the process or change it.

The mechanism FERC chose, the Section 206 show-cause order, is stronger than a request and weaker than a mandate, and the distinction matters. Rather than prescribing a single national rule, the Commission put the burden of proof on each operator: explain, within 60 days, why your current tariff is still just and reasonable for an era of gigawatt loads, or file something better. That structure produces six tailored answers rather than one, which fits a country where PJM’s congestion problem and CAISO’s are not the same problem, but it also guarantees a wave of contested tariff filings through the second half of the year, each of which becomes its own fight over who pays for what.

The Cost-Shift Fight Is the One That Sets Prices

Of the five reform areas, two are operational speed and three are about money and structure, and the money ones are where the hosting industry’s economics are decided. The second reform area, preventing cost shifts and requiring transparency on transmission costs, is the political core of the entire data-center power debate. When a large new load triggers grid upgrades, someone pays for the new wires and substations, and the unresolved question is whether that cost falls on the data center that caused it or is spread across every household and business on the system. For a hosting or cloud operator, the answer flows straight into the cost base: if large loads must fund their own grid upgrades, the capital cost of new capacity rises and is passed through to whoever rents the compute; if the cost is socialized, capacity is cheaper to bring online but the political license to build it erodes.

The third and fifth reform areas, co-location and behind-the-meter generation and the study process for generation serving “electrically proximate” loads, are FERC writing rules for the arrangement the largest buyers have already started defaulting to. Unable to wait for the grid, hyperscalers and their partners have been pairing data centers directly with dedicated generation, on-site power that sidesteps the interconnection queue entirely. These deals have been proliferating faster than the rules governing them, and contested cases have turned on whether a data center drawing from a co-located plant still owes its share of the grid it relies on for backup. FERC putting “electrically proximate loads” in a show-cause order is the Commission acknowledging that the workaround has outgrown the rulebook, and the rules it produces will determine whether co-location remains the fast lane for the best-capitalized buyers or gets re-priced to carry its true system cost.

Faster Interconnection Is Coming. Cheaper Is Not Guaranteed.

The honest read for the hosting industry is that this is upstream, structural, and slow-acting, which makes it easy to ignore and expensive to have ignored. No price changes this quarter because of it. But the interconnection rules these six operators file over the next four months will shape the supply and the cost of the data-center capacity that everyone from the hyperscalers down to the regional managed host ultimately resells. Faster, more predictable interconnection means more capacity coming online closer to schedule, which over a multi-year horizon is the only real relief for the compute-supply tightness that has repriced everything from GPU instances to, by extension, the VPS and shared-hosting markets that sit downstream of the same power and hardware costs. The cost-shift decision determines whether that new capacity arrives cheap or expensive. And the co-location rules determine whether the advantage of building your own power stays concentrated among the few operators large enough to sign gigawatt energy deals, which has direct implications for competitive structure: if behind-the-meter power remains the fast lane and only hyperscalers can use it, the gap between them and everyone else widens at the level of physics.

The scope limitation is the part to watch with skepticism. FERC reaches the organized markets but not the roughly one-third of the country in non-RTO regions, where it could only encourage transmission owners to follow suit. Those regions, much of the Southeast, include significant data-center development, and a reform that speeds interconnection in PJM while leaving a patchwork everywhere else will redirect siting decisions as much as it accelerates them. For operators making location choices, the regulatory map just became another input alongside power price and tax incentive. The Commission has set the clock; the six filings due by late summer, and the contested proceedings that follow them, are where the actual rules, and the actual costs, get written.