TL;DR
X reads gigawatts of data-center requests as a settled AI boom or a coming blackout; ERCOT and PJM screen out the speculative, and one project can sit in two queues at once.
MSM Perspective
MSM such as Bloomberg and the Texas Tribune report the headline gigawatt totals, less the screening that culls speculative requests.
X Perspective
X treats interconnection queue totals as a forecast — proof of an unstoppable AI boom or an imminent grid failure.
A number is loose in the energy discourse again, and it is still being read as a prophecy instead of an application.
The paper argued on June 27 that ERCOT's large-load queue counts requests, not commitments — a pile of data-center applications, most of which never get built. A day later the more useful fact is who does the culling, and where the double-counting hides. ERCOT's large-load integration process screens proposed projects for the studies, deposits, and milestones that separate a serious interconnection from a placeholder before any megawatt is counted as firm. [1] The queue is the inbox; the screening is the edit.
That edit happens in public. ERCOT's Large Load Working Group — a non-voting body that reports to the Technical Advisory Committee and recommends the rules for integrating these loads reliably — is where the standards get argued line by line, in the open, by the people who must keep the lights on. [2] A gigawatt figure shouted on X has skipped that room entirely.
The definitional sloppiness travels because the terms are not standardized across regions, and a developer can file the same speculative campus in more than one queue. PJM, the largest U.S. grid operator, runs its own service-request process with its own thresholds and its own backlog. [3] Add the headline totals across ERCOT and PJM and you double-count ambition — one project, two inboxes, counted twice as destiny.
This is the divergence the paper keeps. X reads the gross queue figure as a settled forecast; mainstream coverage reports the same gigawatt totals with a worried adjective. Both skip the screening layer that does the actual work — the studies, deposits, and milestones that turn a request into a project or quietly drop it. [1][2]
The stakes are concrete. Texas rates, reliability, and transmission build-out depend on which loads are real, because a grid cannot plan against a number that includes projects no one intends to finish. A planner who treats every request as a commitment over-builds and bills households for it; one who dismisses the queue entirely under-builds and risks the shortage. The integration process exists to find the line, and the working group exists to keep that line honest. [1][3]
A queue figure measures interest, not concrete poured. Until a feed can say how many of those gigawatts cleared screening, signed milestones, and posted deposits — and how many were filed twice — it is quoting an inbox and calling it a forecast. [2]
-- DARA OSEI, London