FERC Pressures Grid Operators to Unclog AI’s Power Bottleneck

TL;DR

• FERC issued Section 206 “show cause” orders to all six major U.S. regional grid operators on June 18, 2026, forcing them to defend or reform interconnection rules for AI data centers.
• The orders explicitly target rules that slow large-load customers — specifically AI data centers — from connecting to the grid, with mandates to balance speed, reliability, and consumer cost protection.
• The move effectively kicks off a national rulemaking on how fast and cheap AI infrastructure can plug in, with direct stakes for OpenAI, Google, xAI, Anthropic, and every hyperscaler racing to secure power for compute clusters.
• Utilities and consumer advocates are expected to fight back over who pays for grid upgrades and whether expedited AI interconnections shift costs onto residential ratepayers.

FERC Targets Six Regional Operators With Tailored Orders

On June 18, 2026, the Federal Energy Regulatory Commission issued Section 206 “show cause” orders to all six major U.S. regional grid operators — excluding Texas, which operates its own isolated grid. The orders require each operator to either defend their existing interconnection frameworks or propose reforms to allow large-load customers, specifically AI data centers, to connect to the power grid faster while maintaining reliability and controlling consumer costs.

The orders invoke Section 206 of the Federal Power Act, a regulatory hammer FERC swings when it believes existing rules are unjust, unreasonable, or unduly discriminatory. Translation: FERC thinks the current system is broken for AI loads, and it’s not asking politely for a fix.

Each of the six operators — covering regions from the mid-Atlantic to the Midwest to California — received tailored orders, meaning FERC isn’t applying a one-size-fits-all template. It’s targeting specific bottlenecks in each region’s queue and cost allocation practices.

Why AI Data Centers Are Strangling Interconnection Queues

Over the last two years, AI-driven data center demand has choked U.S. interconnection queues with multi-year delays for large loads. Regulators have been warning that existing rules were designed for traditional generation assets — wind farms, solar plants, natural gas peakers — not clusters of 100 to 1,000+ megawatt data center loads dominated by AI workloads.

The problem? A typical wind farm ramps up gradually and sells power into the grid. An AI data center is a giant, always-on electricity vacuum that sucks down power 24/7 with minimal flexibility. Grid operators treat them like generation projects in reverse, which means they get stuck in the same glacial interconnection process designed for power plants.

And the queues are long. Some operators have gigawatts of AI load applications sitting in limbo, waiting for grid studies that can take two to four years. By the time the study clears, the AI company has either moved to another region or the competitive window has closed.

FERC’s Gambit: Speed Up AI Without Blowing Up the Grid

FERC’s orders walk a tightrope. The agency wants AI data centers connected faster, but it can’t ignore reliability or dump grid upgrade costs onto residential ratepayers who didn’t ask for a new AI cluster next door.

The move raises real concerns about whether expedited interconnection for AI data centers will shift grid upgrade costs onto other ratepayers and whether regulators can simultaneously meet decarbonization, reliability, and AI growth objectives. Utilities and consumer advocates are expected to push back hard on cost allocation and reliability risk.

Here’s the tension: if an AI data center wants to plug in 500 MW of load, someone has to pay for the transmission upgrades, substation expansions, and voltage support equipment. Traditionally, the entity requesting the interconnection foots the bill. But if FERC pressures operators to speed things up, there’s a risk that grid operators socialize some of those costs across all ratepayers to avoid scaring off AI investment.

I think FERC is betting it can thread the needle — create a fast lane for AI loads without turning this into a subsidy fight. But that’s a political minefield. Consumer advocates will argue that residential customers shouldn’t subsidize OpenAI’s next training cluster, and they’ll have a point.

Think of it like adding a new on-ramp to a highway. If the highway can handle the traffic, great — charge the developer a connection fee and move on. But if the new on-ramp requires widening three miles of highway, who pays? The developer, or every driver who uses the road?

Cloud Giants and Hyperscalers Are Watching Every Word

Cloud providers and hyperscalers — including those behind OpenAI, Anthropic, Google, xAI, and DeepSeek — are all competing to secure low-cost, reliable power for AI clusters. FERC’s actions will shape regional advantages and may accelerate investment in grid-enhancing technologies, nuclear, and renewables near key AI hubs.

This isn’t abstract regulatory theater. Where AI clusters get built depends entirely on where power is cheap, abundant, and fast to connect. If FERC’s orders unlock faster interconnections in the mid-Atlantic or Midwest, those regions instantly become more attractive than California or the Pacific Northwest, where queues are longer and power is pricier.

Google, Microsoft, and Amazon are already scouting sites near nuclear plants and hydroelectric dams. xAI reportedly signed a deal to co-locate near a Texas gas plant. If FERC’s orders tilt the playing field toward regions with cooperative grid operators, we’ll see a land rush.

And it’s not just about location. The orders could accelerate investment in grid-enhancing technologies — think advanced transmission software, dynamic line ratings, and battery storage that smooths out AI load spikes. If grid operators can prove they can handle AI loads without massive infrastructure builds, they win the AI data center sweepstakes.

What Happens Next in the FERC Rulemaking Process

The six regional operators now have a deadline to respond to FERC’s show cause orders. They can either defend their current rules as just and reasonable, or they can propose reforms that speed up AI interconnections without compromising reliability or cost controls.

Expect utilities to push back. They’ll argue that AI data centers are fundamentally different from traditional loads and that rushing interconnections creates reliability risks. They’ll point to grid stress events — summer heat waves, winter freezes — and argue that adding gigawatts of inflexible AI load without proper studies is reckless.

Consumer advocates will pile on with cost allocation arguments. If FERC forces operators to speed up AI interconnections by socializing upgrade costs, residential ratepayers will sue. And they’ll win public opinion, because nobody wants their electric bill subsidizing a tech company’s AI ambitions.

But FERC has leverage. Section 206 orders put the burden of proof on the grid operators. They have to prove their current rules are reasonable, not the other way around. If they can’t, FERC can impose reforms unilaterally.

The timeline matters too. If FERC moves fast, we could see new interconnection rules by late 2026 or early 2027. That would unlock a wave of AI data center construction just as the next generation of training clusters — think GPT-5 scale or beyond — comes online.

Source: buildfastwithai (summarizing FERC orders)