Washington Puts the Grid on Notice
The Federal Energy Regulatory Commission has done something unusually direct. Under Section 206 of the Federal Power Act, it issued tailored show cause orders to each of the six regional grid operators under its jurisdiction: PJM, MISO, SPP, CAISO, ISO New England, and NYISO. The orders instruct every one of them to either justify why their current rules for connecting large energy users remain just and reasonable, or file tariff changes that address the problems the Commission has identified. This is not a gentle request for comment. A show cause order flips the burden of proof onto the operators.
The timeline is aggressive by regulatory standards. Each operator and its transmission owners have 60 days to defend or reform their tariffs, and within 30 days they must submit a detailed report describing how they intend to ensure adequate generation is available to serve both existing and new large loads. FERC framed the effort as an aggressive, targeted action to speed large load integration, language that signals impatience with an interconnection process widely seen as too slow for the moment.
Why Data Centers Forced the Issue
The subtext of these orders is artificial intelligence, and FERC is not hiding it. The Commission named data centers, manufacturing facilities, and other large energy users as the drivers, but it is the gigawatt-scale AI campuses that have broken the old assumptions. Interconnection rules were written for a world where new large loads arrived occasionally and grew incrementally. AI has replaced that cadence with requests for hundreds of megawatts to multiple gigawatts, arriving faster than transmission planners can study them and clustered in regions where the grid was never sized for that draw.
We have written before about how power has become the binding constraint on AI expansion, and this is the regulatory system catching up to that reality. When a single hyperscale campus can request as much electricity as a mid-sized city, the question of how and whether it connects stops being a routine engineering matter and becomes a question of grid reliability, cost allocation, and public interest. FERC is effectively conceding that the existing framework cannot absorb this demand gracefully and demanding that the operators fix it.
The Reforms on the Table
The orders sketch several categories of reform. They call for more efficient transmission service application and study processes, including consideration of alternative transmission technologies that could unlock capacity faster than new lines. They demand protections against cost shifting, so that the enormous transmission costs associated with serving a data center do not quietly land on ordinary ratepayers. And they push operators to accommodate co-location arrangements and behind-the-meter generation, the increasingly common pattern of building power directly alongside compute.
Perhaps the most consequential thread is the emphasis on new services for flexible large loads. A data center that can curtail or shift its consumption during grid stress is a fundamentally different proposition than one that demands firm, uninterrupted power at all times. FERC wants tariffs that reward flexibility, which would let operators integrate large loads faster by treating their willingness to bend as a grid asset. For AI operators, that reframes demand response from a nuisance into a lever for getting connected sooner.
What It Means for Infrastructure Leaders
For the executives planning AI capacity, these orders confirm that power and interconnection terms now sit at the center of infrastructure strategy, not at its periphery. The old mental model, in which a company chose a site and then arranged electricity, has inverted. Increasingly the availability of firm power and a favorable interconnection queue determines where compute can be built at all, and the regulatory posture of the local grid operator becomes a due-diligence item as important as land, fiber, or water.
The flexibility theme carries a direct operational implication. Enterprises and hyperscalers that can architect workloads to tolerate curtailment, shifting non-urgent training or batch inference to periods of grid slack, may find themselves at the front of the interconnection line rather than the back. That is a meaningful design consideration, because it turns a capability that engineering teams already possess, the ability to schedule and pause workloads, into a lever for faster access to the scarcest resource in the industry.
A Signal That the Buildout Has Limits
Zoom out and the FERC action is a marker of a broader inflection. For two years the AI infrastructure story has been told in the language of ambition: bigger campuses, larger capital commitments, more gigawatts. These orders introduce the countervailing language of constraint, the recognition that the grid is a shared, physical system with finite headroom and that unlimited demand cannot be waved through without consequences for reliability and cost. The regulator is asserting that the buildout must be reconciled with the grid, not simply imposed on it.
How the six operators respond over the coming weeks will shape the economics of AI infrastructure across the country. Tariffs that reward flexibility and streamline studies could accelerate the buildout in disciplined ways. Rules that fail to prevent cost shifting could provoke public backlash that slows everything down. Either way, the message to the industry is unambiguous: the era of treating electricity as an afterthought is over, and the companies that internalize grid reality into their planning will build faster than those that do not.
