On June 18, the Federal Energy Regulatory Commission unanimously ordered the six regional power grid operators that cover most of the United States to justify or rewrite the rules governing how large electricity users like AI data centers connect to the grid. The action gives data center developers a government-backed pathway to power that did not exist two weeks ago.

This is not an abstract policy shift. Energy has quietly become one of the most concrete constraints on AI deployment at scale, and Washington just made it a national priority.

What FERC Actually Did

FERC issued tailored show-cause orders to PJM Interconnection, Midcontinent Independent System Operator (MISO), Southwest Power Pool, California Independent System Operator, ISO New England, and New York Independent System Operator — the transmission organizations covering most of the contiguous US grid.

Each operator received two deadlines. They have 30 days to submit a report on available generating capacity in their region. They have 60 days to either defend why their existing electricity rate structures remain adequate for large-load customers, or propose revisions that would accommodate the scale and speed that AI campuses actually require.

Rather than launching a Notice of Proposed Rulemaking, which typically takes years to finalize, FERC used Section 206 of the Federal Power Act to issue customized orders to each operator. The practical effect is speed. Grid operators are now on a 60-day clock to respond, rather than waiting through a multi-year rulemaking process.

Data centers will pay their own interconnection costs under the framework, keeping the financial burden off ratepayers while removing the procedural delays that have made grid connection one of the longest lead-time items in data center development.

Why Energy Became the Bottleneck

AI infrastructure has scaled faster than the grid was designed to handle. An AI campus seeking several hundred megawatts is not unusual in 2026. Requests for gigawatt-scale connections are increasingly common. The transmission systems, generation queues, and planning rules that govern those connections were written for much smaller and more predictable load profiles.

The result has been interconnection queues that stretch years out. Developers trying to build data centers to support AI workloads have faced the same problem as renewable energy developers: a technically viable site, a willing operator, and no realistic path to getting power at the scale and speed they need.

The delays have had downstream effects. When data center construction slows, so does the rate at which new AI computing capacity comes online. That shows up in model availability, inference pricing, and the practical ability of businesses to access frontier AI at reasonable cost. The constraints on the infrastructure layer are not invisible to the application layer.

The Policy Context

Energy Secretary Chris Wright directed FERC in October 2025 to consider reforms aimed at timely interconnection of large loads. FERC’s June 18 action represents that directive translating into concrete regulatory movement.

The framing from both the administration and FERC is national competitiveness. The argument being made explicitly is that AI infrastructure capacity is now a strategic asset in the same category as manufacturing capacity or semiconductor production. Allowing procedural gridlock to slow AI data center development is, in this framing, a competitive risk that cannot be treated as a routine regulatory matter.

What This Means for Business

Most businesses deploying AI are not building their own data centers. But they are directly affected by the infrastructure constraints FERC is addressing. A few things to track:

Pricing stability. One of the less-discussed factors in AI model pricing has been compute scarcity. As more capacity comes online faster, the pricing pressure on frontier model inference should ease. That is not guaranteed, and demand may continue to outrun supply, but removing a structural bottleneck helps.

Geographic concentration risk. Current AI data center capacity is concentrated in a small number of locations where grid connections have been feasible. Faster interconnection across more grid regions means more distributed capacity, which improves resilience for businesses whose workloads depend on specific availability zones.

Build timelines. For enterprise teams considering on-premise or co-location AI infrastructure, interconnection queues have been a planning uncertainty. Grid operators now have a federal mandate to address those queues explicitly, which changes the risk profile of large infrastructure commitments.

Longer horizon. FERC’s action sets a process in motion, not an immediate outcome. Grid operators have until mid-August to respond, and actual tariff changes will take additional time to implement. The signal is important, but do not expect the data center market to transform in the next quarter.

The energy constraint on AI has been one of the more credible arguments for why AI deployment cannot simply scale indefinitely at current trajectory. FERC’s June 18 action does not eliminate that constraint, but it is the first federal regulatory move to treat it as a problem requiring urgent attention rather than a market dynamic that will sort itself out.

For businesses planning AI deployment strategy over a one-to-three year horizon, grid capacity is now a regulated priority rather than an unmanaged bottleneck. That is a meaningful change in the infrastructure picture.

If you are thinking through what AI infrastructure looks like for your business, whether that is cloud-based AI agents, custom applications, or embedded AI in your operations, talk to the Enterprise DNA team. Understanding what sits below the application layer is often where the real deployment decisions get made.