Developers racing to build the data centers that power artificial intelligence are running into a wall that no amount of capital spending can fix: the electric grid cannot connect new facilities fast enough. The Federal Energy Regulatory Commission has taken direct action on co-location disputes tied to AI-driven power demand, and its interconnection reform orders target queue backlogs that have stalled projects for years. The result is a widening gap between the money flowing into AI infrastructure and the physical ability to plug it in.
Grid bottlenecks stall data centers even as AI budgets surge
The tension is straightforward. Companies have committed record sums to build data centers, but getting those facilities connected to the power grid requires passing through an interconnection queue that is badly overloaded. FERC has acknowledged that queue volume and limited transmission capacity are the two structural forces behind these delays, according to the agency’s interconnection explainer. Every month a project sits in that queue, construction timelines slip, costs climb, and the AI services those facilities are meant to support stay offline.
FERC’s Order No. 2023 and its follow-up, Order No. 2023-A, were designed to process interconnection requests faster by imposing new study deadlines and financial commitments on applicants. The reforms aim to clear out speculative projects and give serious developers a shorter path to energization. But the rules treat all grid-connected generation and load requests through the same general framework, which creates a specific problem for data center operators pursuing co-location arrangements with existing power plants.
Co-location, where a data center sits behind the meter at or near a generating station, has become a preferred strategy for companies that want to bypass years-long queue waits. The approach lets a facility draw power directly from a generator without needing a separate grid interconnection. That shortcut, however, has triggered disputes over whether co-located loads should still pay for transmission upgrades that benefit the broader system, and whether they create reliability risks for other customers on the grid.
FERC’s co-location orders and the PJM-Constellation dispute
FERC moved to address these disputes head-on. The commission ordered further action on co-location issues after determining that AI and data center load growth is creating serious market-rule problems. The proceedings involve PJM transmission owners and a complaint filed by Constellation, the large nuclear generator operator that has pursued co-location deals to serve hyperscale data center customers.
The core question FERC is trying to resolve is how co-located data centers should be treated under existing tariff rules. If a data center draws power behind the meter but can still pull from the grid during outages or demand spikes, other ratepayers may end up subsidizing transmission infrastructure that serves the data center indirectly. PJM’s transmission owners have argued that current rules do not adequately account for this scenario, and FERC’s show-cause proceedings signal that the commission agrees the existing framework needs revision.
This regulatory uncertainty has a direct effect on project timelines. Developers considering co-location cannot move forward with confidence until FERC clarifies cost-allocation responsibilities. Projects that entered planning as co-located facilities now face the prospect of being reclassified or required to go through the standard interconnection queue after all, adding years to development schedules. That risk is especially acute for AI-focused campuses that are designed around extremely high, relatively constant loads, which can require substantial network upgrades if they are treated as traditional grid customers.
Interconnection reform meets AI’s growth curve
Order No. 2023 is meant to alleviate some of these pressures by reshaping how regional transmission organizations and utilities study and approve new connections. The rule pushes grid operators toward cluster studies, standardized modeling assumptions, and firm deadlines, with penalties for missed milestones. In theory, that should help both generators and large loads secure capacity more predictably, reducing the mismatch between AI project announcements and actual energization dates.
Yet even a more efficient queue cannot fully solve the problem created by AI’s growth curve. Data center developers are planning multi-gigawatt campuses that can come online in phases, ramping demand faster than transmission lines and substations can be reinforced. Where transmission is already constrained, grid operators may struggle to accommodate that scale of load without extensive upgrades, regardless of how quickly studies are completed under FERC’s new framework.
That tension explains why co-location remains attractive despite its regulatory complications. By pairing with existing generators, especially nuclear or large thermal plants with steady output, data centers can in principle secure a dedicated power source that is already integrated into the grid. The plant continues to inject power at its existing interconnection point, while the data center effectively becomes an on-site customer. From a developer’s perspective, that can look like a way to lock in capacity without waiting for a new queue position to clear.
What comes next for AI-driven grid policy
FERC’s twin focus on interconnection reform and co-location disputes suggests that AI demand is now a central driver of federal grid policy. The commission’s orders signal to regional grid operators that they must adapt their tariffs to account for large, flexible loads that may sit physically close to generation but still rely on the broader transmission system for reliability. How those rules are written will determine whether co-location remains a viable shortcut or becomes just another path into an already crowded queue.
For developers, the message is that power strategy can no longer be an afterthought in AI infrastructure planning. Securing capacity, understanding potential cost responsibilities for transmission upgrades, and tracking evolving FERC guidance are now as critical as choosing chip suppliers or cloud partners. Until the regulatory picture stabilizes and transmission investment catches up, the limiting factor for AI expansion will not be algorithms or capital-it will be the wires that connect data centers to the grid.
Warren Cohen is a finance writer based in Phoenix, Arizona, covering personal finance topics including credit, banking, and beginner investing. He earned his degree in business administration from Arizona State University and began his career working in consumer finance, where he gained direct experience with lending and credit systems. He now writes for personal finance websites and fintech platforms, focusing on clear, practical content that helps readers make informed financial decisions.
