Interconnection studies and queues

For most data center projects since 2023, the question has not been what power costs but when it can be connected. Utilities that once connected a 20 MW facility within a year or two now face requests for hundreds or thousands of megawatts at single sites, totaling many times their existing peak demand. Each request must be studied to determine what transmission and generation additions are needed, who pays for them and how long they will take. This lesson explains how large loads and generators are connected to the grid, how interconnection studies work, why queues have grown so long, and what developers can do to secure power on a predictable schedule.

Load versus generator interconnection

Connecting a large load and connecting a generator are different processes with different rules:

• Generator interconnection to the transmission system is regulated by FERC through each transmission provider's open access transmission tariff (OATT). Order 2003 standardized procedures, and Order 2023 (2023) reformed them substantially.
• Load interconnection has traditionally been handled by utilities under state regulation, with retail service and transmission connection governed by utility tariffs and state commission rules. FERC began asserting a role in large-load interconnection to the interstate transmission system in 2025 and 2026.

For a data center, both matter: the load connection determines when the campus can draw power, and generator interconnection determines how quickly the new supply needed to serve it (and any co-located or contracted generation) can come online.

The study process

Whether for a load or a generator, interconnection studies follow a similar sequence of increasingly detailed analyses:

• Pre-application or scoping: the customer describes the project, location, size and ramp schedule; the utility gives preliminary information about nearby facilities and capacity.
• Feasibility or preliminary study: steady-state power flow analysis (ELE-380) identifies thermal overloads and voltage violations under normal and N-1 contingency conditions.
• System impact study: detailed power flow, short-circuit, stability and sometimes electromagnetic transient studies under many scenarios and contingencies, identifying the required network upgrades.
• Facilities study: engineering design and cost estimates for the connection facilities (substation, lines, protection, metering) and network upgrades.
• Agreement: an interconnection or connection agreement specifying the facilities, costs, security, schedule, operating requirements and, for loads, the contracted demand and ramp.

Each stage takes months; the whole process for a large load or generator commonly takes two to four years before construction of upgrades even begins. Upgrades themselves may take years more: a new 500 kV line can take a decade from conception to energization, while large power transformers have had lead times of two to four years.

What studies look for

A large load affects the grid in several ways that studies must evaluate:

• Thermal capacity: whether lines and transformers can carry the additional flow under normal conditions and after contingencies. A 500 MW load at a single point can overload multiple circuits.
• Voltage: large loads depress voltage, especially after contingencies; reactive power support (capacitor banks, STATCOMs) may be needed.
• Resource adequacy: whether there is enough generation to serve the load at peak, which in many regions is now the binding constraint rather than transmission.
• Stability and ride-through: data center loads are dominated by power electronics (UPS rectifiers and server power supplies). When a fault depresses voltage, many UPS systems transfer to battery within milliseconds, removing their load from the grid. In July 2024, a 230 kV fault in Virginia caused about 1,500 MW of data center load to transfer to backup simultaneously, creating a sudden surplus that operators had to manage. NERC has since identified large-load ride-through behavior as a reliability risk, and system operators increasingly require data centers to remain connected through specified voltage and frequency disturbances.
• Power quality: harmonics and, for AI training clusters, rapid load swings of tens of megawatts as computation synchronizes across thousands of GPUs.

Why queues grew

Generator queues grew from about 500 GW in 2014 to more than 2,500 GW at the end of 2023, more than the entire installed US fleet, mostly solar, storage and wind. Most projects never get built: historically, only about one in five projects requesting interconnection reached commercial operation, and the median time from request to operation grew to about five years. Large-load requests grew even faster: ERCOT reported more than 200 GW of large-load requests by late 2025, against a peak demand of about 85 GW, and several utilities reported requests exceeding their entire system peak.

Many requests are speculative: developers file at multiple sites, hoping to secure capacity at one, and utilities have no way to tell which are real. This inflates queues, slows studies for everyone and makes load forecasts unreliable, which in turn affects resource planning and rates.

Federal reforms

FERC Order 2023 replaced the "first-come, first-served" serial study of generators with "first-ready, first-served" cluster studies. Key features:

• Cluster studies: requests received in a window are studied together, sharing upgrade costs.
• Readiness requirements: site control, study deposits that increase at each stage, and commercial readiness evidence.
• Withdrawal penalties: projects that drop out forfeit part of their deposits, discouraging speculative requests.
• Study deadlines: transmission providers face penalties for late studies.

For large loads, the Department of Energy directed FERC in October 2025 to consider a rulemaking asserting jurisdiction over the interconnection of loads above 20 MW to the interstate transmission system, with standardized procedures, study deposits and the option for loads that agree to be curtailable, or that bring their own generation, to be studied faster. In June 2026, rather than a single national rule, FERC issued orders to each RTO and ISO requiring them to show that their tariffs have clear, consistent large-load interconnection rules or to propose reforms covering study processes, cost protection for other customers, co-location and behind-the-meter generation, and new transmission services for flexible loads. Responses and filings were due in late 2026, so details will continue to change. In December 2025, FERC had already required PJM to create three new transmission services for co-located loads.

State responses

States and utilities have adopted large-load tariffs that require commitments from data centers:

• Minimum demand charges: AEP Ohio's data center tariff, approved in 2025, requires new data centers above 25 MW to pay for at least 85% of their contracted demand for up to 12 years, even if they use less; after it was proposed, AEP's data center queue shrank from about 30 GW to 13 GW.
• Texas SB 6 (2025): loads of 75 MW and above must disclose duplicate requests elsewhere in Texas, pay study fees, show site control and provide financial security; ERCOT can curtail them in emergencies, and the PUCT is developing interconnection standards.
• Other states, including Georgia, Virginia, Indiana and Pennsylvania, have adopted or are considering similar tariffs with exit fees, collateral and long contract terms, aiming to protect residential customers from paying for infrastructure built for data centers that never arrive.

Cost responsibility

Who pays for the facilities needed to connect a large load is one of the most contested questions in the industry:

• Direct connection facilities (the customer's substation, the tap line) are normally paid by the customer.
• Network upgrades (reinforcing shared lines and substations) may be paid by the customer, by all customers through rates, or split; rules vary widely.
• Generation needed to serve the load is paid by all customers through rates or capacity markets, unless the customer brings its own supply.

Many utilities use a revenue credit approach: the customer pays for facilities up front or through contribution in aid of construction (CIAC), credited by expected revenue, with minimum bills or collateral guaranteeing that revenue.

Strategies for securing power

Data center developers use several approaches to obtain power faster and more reliably:

• Early engagement: file requests years ahead, share accurate load forecasts and ramp schedules, and build relationships with utility planners.
• Site selection for available capacity: target locations with existing transmission headroom, retired generation sites with existing connections, or areas with surplus generation (DCE-410).
• Phased ramps: request capacity in stages that match the utility's ability to build, with firm early phases and options for later ones.
• Flexibility: agree to curtail load during system emergencies or peak periods in exchange for faster connection. Studies suggest that if large loads could curtail for a fraction of a percent of hours, existing grids could connect tens of gigawatts more load without new capacity.
• Bring your own generation: on-site or co-located generation (Lessons 7 and 10) can serve the load initially ("bridging") or permanently, reducing the burden on the grid.
• Fund upgrades: offer to pay for network upgrades or provide security, making the request more credible and protecting other customers.
• Buy existing positions: acquire sites or projects with existing connection agreements or queue positions.

Lesson 4 examines how flexibility can unlock capacity sooner, and Lesson 9 the contract terms, such as minimum bills and collateral, that come with large-load connections.