Batch Zero, as announced and as filed

A reconciliation is the short format here. Two public documents describe the same process and disagree. The table is the piece.

9 September 2026

For anyone who does not follow this

The continental United States runs on three separate power grids. Eastern, Western, and Texas. Texas is its own on purpose. Federal law reaches electricity that crosses state lines, so Texas utilities never crossed one. Keep the wires inside the state and Washington, DC stays out of the room. ERCOT operates that grid for about ninety percent of the state's load, and it owns neither wires nor power plants. It is closer to an air traffic controller: it balances supply against demand and decides who is allowed to plug in.

For a century the demand side was the easy half of that job. No single customer was large enough to need studying. A whole suburb might draw 100 megawatts.

A single AI campus can want a thousand. That is one address pulling more than one percent of the entire grid. So before anyone connects, someone must run a study: drop this much load here, and do the lines overheat, does the voltage hold, what happens when the nearest generator trips. Running those one at a time, in the order they arrived, never finishes.

Batch Zero is the fix. Take every request of 75 megawatts3 and above, study them together as one problem, publish one plan.

The scale

Peak demand from ERCOT's all-time records.1 Queue figures from ERCOT's April update to the Texas Senate2 and its June announcement.3 Bars are to scale.

Data centers were about 87 percent2 of that queue in April and nearly 89 percent3 in June. It grew 28,000 MW in eleven weeks, and it is asking for roughly four point eight times the largest load Texas has ever carried.

The reconciliation

Both rows are ERCOT's own publications, six weeks apart.34 ERCOT states that all peak demand records are unofficial until final settlements occur.1

Two documents from the same organization, six weeks apart. The first milestone of the process slipped, four days before it was due. No new date was offered. The completion date stayed exactly where it was.

What this is not

It is not a claim that Texas is four hundred gigawatts short. A queue is a list of requests, and a request costs a developer very little. One project can be filed at several sites, and places can be held speculatively. Nobody yet knows what fraction of that 438,000 MW is real.

That uncertainty is the finding rather than a footnote to it. The state cannot presently say how much of its announced capacity is a project, and the study built to answer that has slipped with no new date. What follows is narrower and harder to argue with: every Texas energization date announced for 2027 rests on a transmission plan that has not been written.

The pattern is familiar. At London's Crossrail the National Audit Office found management still planning against a December 2018 opening while the milestones beneath it were missed.5 The difference is what binds. Crossrail's constraint was information, which was internal, which is why replanning fixed it: attainment ran near 30 percent before the August 2020 replan and near 90 percent after, same contractors, same railway.5

A grid queue is internal to nobody. No developer replans their way out of it. When the binding flow is supplied by a shared system upstream of every project, they do not fail one at a time. They fail together.

The full essay: the railway that existed before it worked.

by Ahmed Schrute