The federal government just gave America’s largest power grid operators a 60-day ultimatum: prove your interconnection rules can handle the AI data center surge, or rewrite them. The stakes are not hypothetical. Hyperscalers have committed hundreds of billions of dollars to AI infrastructure that cannot run without grid access, and the queue to plug in is getting longer by the month.
The Federal Energy Regulatory Commission on June 18 issued tailored show-cause orders to six regional grid operators covering more than 200 million Americans across 30-plus states and the District of Columbia. PJM Interconnection, MISO, Southwest Power Pool, CAISO, ISO New England, and NYISO must now either defend their existing interconnection frameworks or propose concrete reforms that speed up grid access for large-load customers, specifically AI data centers drawing 20 megawatts or more.
Why FERC Moved Now
The commission identified a growing bottleneck that threatens to choke the AI infrastructure buildout at the exact moment capital is flooding into it. Interconnection queues for large electricity loads have ballooned as hyperscalers race to bring multi-gigawatt campuses online. Applications are stacking up faster than grid operators can study them, creating a backlog that adds years to project timelines and billions in holding costs.
FERC Chair Laura Swett called the reform a “national priority,” signaling that the agency views grid access speed as a competitive advantage for the United States in the global AI race. That framing puts FERC squarely in the camp that treats AI infrastructure as strategic national capacity rather than routine commercial development.
The orders target five specific reform areas: transmission service application and study processes, cost-allocation mechanisms designed to prevent existing ratepayers from subsidizing data center upgrades, co-location arrangements and behind-the-meter generation, new transmission services for flexible large loads, and processes for studying generation facilities serving electrically proximate large loads.
The Cost Shift That Matters
The most consequential provision may be the cost-allocation mandate. FERC is pushing interconnection expenses directly onto the requesting entities rather than socializing them across all ratepayers. Translation: if Microsoft, Meta, or Amazon want priority grid access for a new AI campus, they pay for the transmission upgrades their facility requires. Existing utility customers do not absorb those costs.
That principle matters because public opposition to AI data centers has been rising fast. A Reuters/Ipsos poll published June 12 found that 57 percent of Americans oppose data center buildouts when told the projects could raise their electricity bills. FERC’s cost-isolation framework is a direct response to that political reality: build fast, but do not ask voters to fund the buildout through higher utility rates.
The policy also reframes the competitive dynamics among hyperscalers. Companies with deeper balance sheets and lower costs of capital gain a structural advantage when interconnection costs fall directly on the applicant. Amazon and Microsoft can absorb multi-billion-dollar transmission upgrades more easily than mid-tier cloud providers or colocation operators, potentially concentrating the AI infrastructure market further.
What Operators Must Deliver
Grid operators face two deadlines. Within 30 days, they must submit reliability reports detailing how they will secure sufficient generation capacity to support the wave of large-load applications without degrading service for existing customers. Within 60 days, they must either justify why their current tariffs remain “just and reasonable” under Section 206 of the Federal Power Act or file proposed reforms.
The orders are tailored rather than blanket. Each operator received requirements calibrated to its specific market structure and queue backlog. PJM, which covers 13 states from Illinois to Virginia and hosts the densest concentration of U.S. data center capacity, faces the most detailed scrutiny. The sheer volume of pending applications in PJM’s territory made it the obvious pressure point.
Texas Stays Outside the Fence
One notable absence: ERCOT, the Electric Reliability Council of Texas. Because Texas operates its own grid outside federal jurisdiction, FERC’s orders do not apply there. That gives Texas a regulatory speed advantage for data center siting, though the state’s grid reliability history during extreme weather events introduces a different kind of risk that enterprise customers must weigh carefully.
The Bigger Picture
FERC’s action lands in the middle of a broader reckoning over who pays for the AI buildout and who benefits. Morgan Stanley estimated this month that hyperscalers will collectively take on $570 billion in AI-related debt. Meta announced a $200 billion data center complex in Louisiana. Amazon signed a multibillion-dollar fiber deal with Corning specifically to support data center connectivity. Meanwhile, Illinois and Ohio have frozen new data center tax incentives amid voter backlash over the gap between promised jobs and actual local benefit.
The commission is threading a needle: accelerate the grid buildout that the AI industry demands while shielding the ratepayers who never asked for it. Whether grid operators can deliver meaningful reforms within 60 days, or whether this becomes another regulatory cycle of extensions and watered-down compliance filings, will determine how fast the next generation of AI infrastructure actually gets built and who bears the cost when it does.