AI Data Center Power Meets a Federal Line
AI data center power has become the defining bottleneck of the artificial intelligence boom, and on June 18 Washington moved to break it. The Federal Energy Regulatory Commission issued tailored show cause orders to all six regional grid operators under its jurisdiction, directing each one to justify or rewrite the rules that govern how data centers and other giant electricity users connect to the system. The action is the bluntest tool the Commission has, and it signals that the federal government now treats AI data center power as core national infrastructure.
The orders were issued under section 206 of the Federal Power Act, the provision that lets FERC declare existing rates and rules unjust and force a change. That legal posture matters. The Commission is not politely requesting comment. It is telling PJM, MISO, the Southwest Power Pool, the California Independent System Operator, ISO New England and the New York Independent System Operator that their current tariffs may already be unlawful, and that they have 60 days to prove otherwise or fix them. For an industry that measures grid reform in years, that is a sprint.
FERC Chairman Laura V. Swett framed the move as both an economic and a security imperative. "We are setting the stage for a resilient, reliable, and forward-thinking grid that empowers communities and safeguards consumers by transforming the way large energy users access the grid," she said, adding that the Commission also wanted to "provide certainty for investors by directing the markets to protect existing deals." The message to hyperscalers was unmistakable. AI data center power is now a strategic asset worth clearing red tape to secure.
What the Six Orders Actually Demand
Each grid operator faces the same two-part deadline. Within 60 days, the operator and its transmission owners must either defend their current tariffs as just and reasonable or file changes. Within 30 days, a much tighter window, each must submit an informational report explaining how it will guarantee enough generation to serve both existing and incoming large loads. The Commission deliberately rejected a single national template, noting that the six markets differ too much in design, geography and stakeholder mix for one rule to fit them all.
Underneath the deadlines sit five categories of reform that every operator must address. They are required to build faster transmission service and study processes, including consideration of alternative transmission technologies. They must prevent cost shifting and create transparency into who bears transmission costs. They have to accommodate co-location agreements and behind-the-meter generation, the arrangements where a data center sits next to its own power plant. They must offer new transmission services for flexible large loads that can ramp demand up and down. And they have to develop a process for studying generators that sit electrically close to those loads.
That last cluster is where the technical fight over AI data center power will be won or lost. Co-location has become the favored strategy for companies that cannot wait years for a grid connection, because pairing a campus directly with a gas plant or a nuclear unit sidesteps the queue. Regulators have worried that those private deals quietly hand grid costs to ordinary customers. By forcing operators to write transparent rules instead of improvising case by case, FERC is trying to legitimize co-location while fencing off the worst of the cost transfers.
The Bottleneck That Triggered Washington
The proximate cause was a directive from the Secretary of Energy, whose Advance Notice of Proposed Rulemaking last October asked FERC to tackle the timely interconnection of large loads, generally defined as demand above 20 megawatts. The deeper cause is arithmetic. Training and running frontier models has turned individual campuses into gigawatt-scale consumers, the equivalent of a mid-sized city arriving on the grid at once, and the serial, first-come-first-served interconnection process was built for a slower world. Thousands of projects have been stuck in study queues for years.
FERC did not arrive at this point cold. Commission staff reviewed more than 3,500 pages of public comment in the large load docket, and the new orders build on a December 2025 order that directed PJM to adopt transparent rules for loads co-located with generation, plus approval of the Southwest Power Pool's High Impact Large Load initiative. The June action knits those one-off interventions into a system-wide standard. Details of the underlying proceeding sit in FERC's large load interconnection docket, and the agency laid out its reasoning in an official news release.
One jurisdiction is conspicuously absent. Texas, whose ERCOT grid operates outside federal authority, is not covered by the orders, even though it has become one of the fastest-growing data center hubs in the country. That gap hands ERCOT a potential speed advantage while the six federal markets renegotiate their rules, and it underscores how fragmented the national approach to AI data center power remains.
Who Pays for the Buildout
The most politically charged thread running through the orders is cost allocation. Connecting a hyperscale campus can require new substations, upgraded lines and additional generation, and the central question is whether the tech company requesting the power or the household down the road ends up paying for it. FERC's insistence on transparency and its explicit aim to prevent cost shifting are a direct response to consumer advocates who warn that ratepayers are subsidizing the AI economy through their monthly bills.
The Commission is trying to thread a difficult needle. Move too slowly and the United States risks ceding the AI infrastructure race to faster jurisdictions abroad. Move too fast without protections and voters absorb the cost of private compute. Swett's emphasis on safeguarding consumers "even as electricity demand and technology accelerates" is an acknowledgment that the political durability of the AI buildout depends on the public not feeling fleeced. Consumer groups have already signaled they will challenge any rulemaking that lacks firm cost-allocation rules, which means litigation is likely no matter what the operators file.
For the hyperscalers, the orders are mostly good news. A faster, clearer path to AI data center power reduces the single biggest constraint on their expansion plans, more binding now than chips or capital. For utilities, the message is more complicated. They gain a mandate to build, but also a federal referee scrutinizing how they price it.
The Stakes for the AI Race
This is the federal government's first system-wide attempt to clear the grid bottleneck choking the AI buildout, and its success is far from guaranteed. The 30-day and 60-day clocks are aggressive by regulatory standards, and operators may respond with defensive filings that defend the status quo rather than embrace reform. If they do, FERC will face the harder task of imposing changes over their objections, a process that can stretch for years and invite appeals.
Yet the symbolism is hard to overstate. Energy regulators have effectively declared that the way America connects large loads is a national priority on par with the chips and models that depend on it. The contest that began in research labs has now reached the most physical layer of the economy, the wires and transformers that move electrons. Whether the country can deliver AI data center power on the timeline the technology demands will shape not just the next product cycle but the geography of where the AI economy is built. For ongoing coverage of how this plays out across markets and policy, follow Blumefield.
The next 60 days will reveal whether the six grid operators treat the orders as a genuine mandate or a box-checking exercise. Either way, the era of treating data center demand as just another connection request is over. Washington has decided that AI data center power is too important to leave to a queue.