Reuters put the first hard national number on the gap between capex committed and capacity actually energized: 700GW+ of large-user grid connection requests, roughly ten times current US data-center consumption.
A grid interconnection request is free to file and carries no obligation to build, so the queue is an option register rather than a demand forecast. Utilities must nonetheless plan generation and transmission against it, which means phantom requests convert directly into real ratepayer-funded capital spending — and the developer who files ten speculative requests to secure one site externalises the cost of the other nine onto the utility's planning base.
Scenarios 2
The gatekeeping spreads and the queue collapses toward something real. Pennsylvania's 20-of-100-plus permit ratio becomes the template other states copy, deposits and readiness tests get attached to requests, and the headline GW number falls sharply without any actual project being cancelled.
Two states are already doing it — Texas via the PUCT audit-and-freeze this map logged on 08-03, Pennsylvania via Shapiro's 08-18 executive order for projects 25MW and up. A third state adopting a readiness test within two quarters is the checkable confirmation, and each adoption should show up as a visible drop in that state's own published queue.
Precedent — The FERC interconnection queue reform of 2023 (Order 2023) did exactly this to renewables: first-ready-first-served replaced first-come-first-served, deposits rose, and speculative megawatts left the queue without reducing built capacity.
The gap persists because it is nobody's job to close it. Developers are rewarded for optionality, utilities earn a regulated return on the capital they deploy against the queue, and states competing for the same projects cannot afford to be the one that adds friction.
The tell is the direction of ratepayer cases. If large-load tariffs and minimum-take provisions spread faster than queue reform does, the system is choosing to socialise the cost rather than to filter the requests. This map already carries the TVA data-center rate effective 10-01 as a dated instance of exactly that response.
Precedent — The 2000-2002 telecom fibre buildout. Announced capacity ran an order of magnitude ahead of realised demand, no participant had an incentive to publish a truthful number, and the correction arrived as write-downs rather than as a gradual repricing of the queue.
Context
Confidence is well-supported rather than plausible because the quantification is now multi-sourced and self-consistent: 700GW+ in the aggregate, 474GW+ in Texas alone matching this map's own 08-03 figure, ~270GW across ten named utilities, and Pennsylvania's 20-of-100-plus.
This is generated reasoning, not a sourced fact — the mechanism and scenarios above are the model's read on what this item could mean for capital flows, tagged with its own confidence rather than stated as settled. Back to Global Capital.