Most of the queue will never be built.

Every ISO and RTO maintains an interconnection queue — an ordered backlog of proposed generators, storage, and hybrid projects seeking a grid connection. Projects move through feasibility, system impact, and facility studies. Capacity shown on utility hosting maps often assumes every queued project will succeed.

Lawrence Berkeley National Laboratory's Queued Up research finds roughly 75% of proposed U.S. interconnection capacity withdraws before commercial operation. In this dataset, the capacity-weighted national withdrawal rate is 61% across 5,660 active projects.

Without interconnection queue prediction, speculative megawatts get treated as firm capacity. A substation that looks full may have headroom once likely withdrawals are removed. One that looks open may be stuck in cluster studies for years.

Large loads usually interconnect through a separate utility process. Deliverable generation-queue capacity is still a strong signal of where grid headroom is likely to exist — a practical input before a data center interconnection study.

A Cox survival model is fit on historical generation-interconnection projects. Each active project gets a withdrawal probability from attributes knowable near queue entry: cluster studies, service type, technology, and the regional base rate.

Those probabilities roll up to likely withdraw MW. Subtracting them from queued megawatts yields adjusted available MW, ranked by the GridSolve Site Score.

The methodology shows regional validation, data sources, and what the model does not claim.

GridSolve covers National — 9 regions. Model validation varies by region. — 4,687 substations across 9 ISO/RTO regions and 48 states. Withdrawal rates are computed inside each region from that region's resolved projects. Predictive strength varies, and each substation carries its region's confidence tier.