The Federal Energy Regulatory Commission on June 18, 2026, launched targeted enforcement action ordering regional transmission organizations and independent system operators to justify or reform their tariffs for connecting data centers and other large energy users to the interstate grid — the sharpest regulatory move yet to force power grids built for smaller loads to accommodate the massive electrical demands of modern AI.

The order is the capstone to a sequence of regulatory actions that began roughly 16 months earlier. In February 2025, the commission voted unanimously to open a formal review of co-location — the practice of siting large data centers directly at power plants to avoid waiting in long connection queues. By October 2025, the Department of Energy had directed FERC to accelerate data center interconnection, opening a formal rulemaking in January 2026. In December 2025, FERC ordered PJM Interconnection — the grid operator serving roughly a third of U.S. electricity consumption — to create new rules and fast-track applications for large loads. FERC committed publicly to finishing a final rule by June 2026, and the June 18 action delivers on that deadline.

What the June Order Actually Does

The June action requires grid operators across the country to either prove that their existing fee structures and connection processes are fair for data centers and other large power consumers, or propose changes to them. The effect is to signal that decades-old connection procedures — designed around smaller, more varied loads like wind farms and gas power plants — now face mandatory review.

FERC is also using automation and artificial intelligence tools inside its own application processing system, aiming to approve more than 60 new interconnection permits by the end of 2026, according to the commission's January 2026 operational outlook. There is an irony here: the agency responsible for regulating the infrastructure that AI requires is now using AI to speed up the paperwork.

The Core Problem

Think of the U.S. electrical grid as a highway system designed in the 1960s. Most exits were built to handle steady traffic from cars and trucks — consistent, predictable flows. Now imagine a single facility that needs to draw as much power as a mid-sized city, 24 hours a day, every day, and that facility wants to plug in at an exit designed for a fraction of that load.

A large AI data center typically requires 100 to 500 megawatts continuously — far heavier and more constant than the loads power plants were originally sized to handle. When you plug that much power demand into a substation serving a regional power hub, it can create cascading effects throughout the grid: fault-current problems, stability issues, and capacity constraints that delay other projects waiting to connect. Co-location — putting the data center right next to the power plant where the electricity is generated — sidesteps that problem. But it raises a legal gray area: if power is consumed on-site before flowing into the public grid, is it subject to federal rules requiring open access to the transmission system.

The 2025 review of co-location was designed to clarify that question, and it directly feeds into the tariff reforms now being ordered across the country.

The pace of regulatory action here is worth noting. FERC historically works on multi-year timelines for rulemaking, so moving from a unanimous vote to formal rules to enforcement orders in 16 months reflects extraordinary political and commercial pressure. The question the commission itself is asking: will these new rule structures hold up as data center designs evolve, or will they need revision. FERC has built review mechanisms into its docket to address that.

For companies with data center projects waiting to connect to the grid, the message from FERC is clear. The regulator is trading the caution of the old process for speed, and power grid operators will follow suit. Applications that looked settled six months ago may be reworked before year-end. That creates uncertainty, but it also signals that the current obstacle — an outdated connection system — is finally being dismantled.

Why This Matters Now

The U.S. grid was not built for this moment. The surge in AI training and inference — tasks that run continuously and draw enormous amounts of power — has collided with a transmission system whose rulebook was written for slower, smaller, and more intermittent loads. That collision is no longer a distant engineering problem; it is unfolding across the country right now, with billions of dollars in infrastructure projects in the queue and no clear mechanism to prioritize or expedite them.

FERC's intervention removes one major bottleneck. How well the new rules work in practice, and how quickly grid operators and power companies adapt to them, will shape the timeline for the next generation of AI data center expansion.