Until the audit and batch results arrive, the requested total is a measure of intent, and the 3.9 GW operating figure is the only number here that was observed.
Researched by an agentic pipeline. Reviewed and approved by the editor before publication.
ERCOT, the Texas grid operator, publishes three large-load numbers that are easily read as stages of one pipeline. They are three different kinds of evidence, and a planner who divides one by another is comparing an instrument reading with a contract status and a statement of intent.123
| Tier | Figure | Basis | As of |
|---|---|---|---|
| Operating | 3,883 MW non-simultaneous; 3,801 MW simultaneous | Observed by ERCOT, built from per-load maxima | Deck dated 13 March 2026; slides labelled "March 2025"1 |
| Approved to energise | 9,042 MW | ERCOT approval to energise; a permission, not a meter reading | Same deck1 |
| Requested | about 410 GW, about 87% data centres | Interconnection requests: stated intent, largely unstudied | 26 March 20262 |
| Requested, later | about 438 GW | Same basis, under the PGRR115-based process | Board deck for 1–2 June 2026; "late May" per trade press35 |
Three of the four rows come from ERCOT, which administers the queue it is counting. The 438 GW date comes from a subscription news service, not from ERCOT's deck, which gives none.35
About 3.9 GW of large load is recorded as operating. ERCOT describes the non-simultaneous figure as the sum of each load's own monthly maximum, which it presents as the approved load it believes is now operational. The simultaneous figure, the highest combined draw in a month, is 3,801 MW. The 82 MW gap shows that individual peaks do not coincide, and it means the larger number overstates what the grid served at any one moment. Both figures fell slightly from February.1
Approval to energise stands at 9,042 MW. The operating figure is 42.9% of that (3,883 ÷ 9,042). That ratio, and the remaining 5,159 MW, are this brief's arithmetic, not ERCOT figures, and they assume the two numbers share a date.1
Requests are another order of magnitude. The 410 GW reported to the Texas Senate was about 87% data centres. By the June board deck the total was about 438 GW, which a trade report says is roughly 28 GW up on the April board meeting. A letter from the governor, as reported in August, puts the queue at about 474 GW, about 90% data centres.2345
Operating is about 1% of the requested total, but the ratio compares a metered figure with an unstudied one. Read it as the scale of the gap in evidence, not as a completion rate.
The operating figure carries an unresolved label. Both operating slides read "March 2025", while the deck is dated March 2026 and its other slides cover the period since February 2025. It may be a typing error, but the text cannot settle it. The brief treats the figure as a March 2026 observation and flags that as an assumption.1
Three further limits apply. The operating figure is ERCOT's alone; no independent re-measurement was found. Whether the underlying readings are metered or reported is not stated in the extractable text. And the charts in the decks, which hold the breakdown by status and zone, were images that could not be read as text.1
One outside figure does not reconcile. A trade magazine says only about 7.5 GW has "actually connected" to the grid, but gives no source for it and no definition of "connected". With a different definition or an earlier date it could be consistent with ERCOT's 3.9 GW; as published it cannot be tested, so it is not used as a measurement here.6
The same article closes with a range of 50 to 150 GW of requests that might become real projects. That range is the author's own hypothetical, not an estimate attributed to anyone, and it should not be quoted as a forecast.6
The requested tier is held back by process rather than by hardware. ERCOT told the Senate that when another project affects the same transmission zone, all earlier studies are invalidated and unenergised projects must be restudied, and that without capacity reservation projects can be pulled into repeated restudies, extending timelines by years.2
The June board deck says the "vast majority" of requests have no studies submitted. A trade report of the same meeting puts the unstudied portion above 300 GW and quotes ERCOT's Jeff Billo saying the data on these requests is thin. That GW figure is the reporter's rendering of a presentation; the deck itself gives no number.35
Billo also expected the headline total to matter less once the current process ended on 10 July 2026 and ERCOT moved to batch studies.35
The batch process has since slipped. After the governor called for a moratorium on new data-centre interconnections on 3 August, ERCOT said it aimed to file an audit of data-centre proposals by 10 December. About 300 data centres of 75 MW or more are in the first batch. ERCOT's Chad Seely said the batch study would not be finished by 9 April 2027 and no new date was set.4
Near-term supply into the operating tier is a projection. ERCOT's stability assessment expects about 3.9 GW of large load to energise in the fourth quarter of 2026. Separately, Utility Dive reports 17 large loads, about 6.6 GW of peak demand, that have passed the stability assessment and are scheduled online later this year, ramping over about five years, but are not yet approved to energise. These are different scopes and horizons, both projected, and the brief does not add them.34
Cancellations are visible but unquantified. ERCOT noted that cancellations in February and early March lowered the queue slightly, with no megawatt figure.1
The fee side is a proposal. Texas regulators proposed in March 2026 a financial security of $50,000 per MW, 20% refundable. Its final status was not checked, and the sources contain no capital-cost or tariff figures.2
It establishes that ERCOT's own records place observed large load below 4 GW, approved load near 9 GW, and requests between 410 and 438 GW as of spring 2026, and that these sit on different evidentiary footings. It establishes that the requested tier is mostly unstudied and that process changes, not demand alone, now govern how much of it can advance.
It does not establish how much of the requested tier is real. Utility Dive reports that experts call a substantial portion speculative or duplicative; no source here measures the share. It does not establish an independent operating figure, the date label on that figure, or any completion rate. It applies to ERCOT's large-load records in 2026, and says nothing about other grids.14
Plan against the tier that matches the decision. Capacity that must exist in the next two years should be sized from the operating and approved figures, and from the 3.9 GW projected for the fourth quarter of 2026. Treat the 410–474 GW range as an indicator of interest in Texas, not as a load forecast or a queue to be cleared in order.
A usable test: any claim built on the requested tier should state how many gigawatts of it have a study, a financial commitment or an energisation date. The sources reviewed do not supply that answer for the requested tier as a whole. Until the audit and batch results arrive, the requested total is a measure of intent, and the 3.9 GW operating figure is the only number here that was observed.
Do not scale the national 52 GW, or Cleanview's 2 GW of gas, into a figure for data-centre storage.
Oversubscription shows demand for allocations; it does not show how much of the allocated time ran jobs.
Treat memory as a priced, volatile input with a public supplier side and a private buyer side, and plan on the private side.