# Grid Interconnection Queues Before a new power plant or a large load such as a data centre can connect to the electricity grid, the grid operator has to study whether the network can handle it and what upgrades are needed. Projects wait in line for those studies. That line is the interconnection queue. ## Why it is slow in the US - Studies are done largely in sequence, and one project dropping out can force restudies for others. - Required transmission upgrades take years to permit and build. - Queues are crowded with speculative requests. Lawrence Berkeley National Laboratory's *Queued Up* reports have shown thousands of gigawatts of proposed generation and storage waiting, with typical waits of around five years from request to operation. Large data centres face similar delays on the load side. ## Units - **Gigawatt (GW):** a measure of power capacity. 1 GW is roughly the output of a large nuclear reactor. A frontier-scale AI campus is now planned in the hundreds of megawatts to multiple gigawatts. - **Capacity added** (e.g. GW added per year) is nameplate capacity. Solar and wind produce well below nameplate on average, so GW added overstates usable firm power. ## Why it matters In the US the binding constraint on AI build-out has shifted from chips to power and grid connection. China adds generation capacity several times faster and plans grid build centrally, so its binding constraint is silicon. Each side is long what the other is short. Related: [[Power Infrastructure]], [[The Watt-Bit Spread and the Future of AI Power Markets]], [[Data Center MoC]], [[AI Capex Super-Cycle]]