Aug 12, 2026 · Research · 8 min

The ERCOT Queue Is Not a Construction Schedule

On June 18, 2026, ERCOT printed a number the industry treats as destiny: 466,497 MW of large-load requests. The same LLWG deck printed 5,700 MW observed energized. That is 1.2 percent. If the queue were a construction schedule, Texas would be building 82 times more large load than it has actually turned on.

Queue-source status: snapshot as of June 18, 2026; source age 70 days as of August 27, 2026. This remains Pallara's latest reviewed ERCOT queue snapshot. It is stale for current-state use. No later reviewed ERCOT queue total replaces 466,497 MW in this article.

It is not a construction schedule. It is a request ledger. A request is not an approval. An approval is not an operating load. The rest of this article is what happens when those three layers get collapsed into one headline.

The sentence vs the source

Most of the damage is done in one sentence. "Texas has 466 gigawatts of data centers in the queue." Four errors live in that line. The megawatts are requests, not plants. The data-center share is not a printed total in the June queue table. The queue is not a schedule. And 466 GW is a rounded restatement of 466,497 MW from one dated deck.

Governor Abbott's August 3 letter used "approximately 90 percent" against an approximate 474 GW request figure. That is a Governor's-letter figure. It is not ERCOT's published queue total. Mixing the two is how 466,497 MW becomes "474 GW of data centers" in a briefing memo.

Three numbers that have to add

The June deck is internally consistent if the layers stay separate. Observed energized (5,700 MW) plus approved-to-energize-but-not-operational (3,200 MW) equals approved to energize (8,926 MW). Slide text on the A2E total prints 8,927 MW. The splits cross-foot to 8,926 MW. One megawatt of rounding. Use 8,926.

graph TB
  Q["Requested
466,497 MW"] --> N["No studies submitted
257,600 MW"]
  Q --> A["Approved to energize
8,926 MW"]
  A --> E["Observed energized
5,700 MW"]
  A --> W["Approved, not yet on
3,200 MW"]

  style Q fill:#e8e8e8,color:#0a0a0a,stroke:#0a0a0a
  style E fill:#e8e8e8,color:#0a0a0a,stroke:#0a0a0a
  style N fill:none,stroke:#0a0a0a
  style A fill:none,stroke:#0a0a0a
  style W fill:none,stroke:#0a0a0a
          

A separate panel prints 3,900 MW as the June 2026 non-simultaneous monthly peak of loads already approved to energize. That is not the 5,700 MW Observed Energized tier. Swap them and the conversion rate moves, the story changes, and the source has been misread. ERCOT's July 17 Monthly Operational Overview later prints 3,966 MW for that same June peak measurement. 3,900 is the LLWG round. 3,966 is the later operational print. Neither is 5,700.

The 28 GW that was not a buildout

A May restatement of "more than 438,000 MW" still circulates. That is the May 20 snapshot (438,595 MW). The current figure is 466,497 MW. The June deck also discloses a pre-Batch-Zero LLIS cleanup: cancelled and duplicate tickets removed, multi-stage tickets combined. Then it restates May at 471,415 MW, which is not what the May deck itself printed.

Batch Zero locked a process. The queue did not move.

On June 18, 2026, the Public Utility Commission of Texas approved ERCOT's Batch Zero package, PGRR145 and NPRR1325. The PGRR145 issue page reads "Approved on 06/18/2026." That locked the process. It did not energize 466 GW. It put filing dates on a calendar. Filing dates are not commercial operation dates.

The PUCT's August 20 order granted all three requested good-cause exceptions. In M-A080326-02, ERCOT says it intends to notify each interconnecting DSP or TSP of its Large Load's conditional classification by August 31. Receipt starts the Planning Guide §9.3.1(2)(a)(i)-(iii) dispute and reconciliation process.

The $100,000 per MW mistake

16 TAC §25.194 remains a proposal in PUCT Project 58481. Comments closed April 17, 2026. The Commission has not adopted a text. The rule has no effective date. Staff's Item 206, filed August 24, 2026, attaches a clean and redline draft of recommended changes and says Staff continues to refine that draft ahead of the September 11 open meeting. A staff draft is not adopted text. A meeting date is not an effective date.

The March proposal's two $50,000/MW figures were already one capital event, not $100,000/MW. Item 206 then struck the non-refundable interconnection fee. Do not write $50,000/MW as the interconnection fee. That fee is not in the current staff draft.

What remains in Item 206 is proposed security. At the intermediate agreement, $50,000 per MW of requested peak demand. At the SLLIA, the greater of $50,000 per MW of contracted peak demand or the costs allocated to the customer for system upgrades. Both are proposed. Neither is a fee.

graph TB
  M["March proposal:
security plus fee"] --> I["Item 206 staff draft:
fee struck"]
  I --> A["Proposed intermediate security
$50k/MW of requested peak"]
  I --> S["Proposed SLLIA security
greater of $50k/MW or
allocated upgrades"]
  X["Struck fee"] -.->|"do not write as $50k/MW fee"| I

  style I fill:#e8e8e8,color:#0a0a0a,stroke:#0a0a0a
  style A fill:#e8e8e8,color:#0a0a0a,stroke:#0a0a0a
  style S fill:#e8e8e8,color:#0a0a0a,stroke:#0a0a0a
  style M fill:none,stroke:#0a0a0a
  style X fill:none,stroke:#0a0a0a
          

Planning Guide study fees are a separate instrument from proposed §25.194 security. Mixing them into an "all-in per MW" number is the same error as treating proposed security as a fee. If a model cannot show the instrument, the evaluation date, and the cited section, it cannot compute the cash event.

A marketing page is not Demand

Campus websites advertise power. That is a public fact about advertised capacity. It is not expected aggregate peak Demand at a single site, not computational share, and not a ride-through finding under NOGRR282 / NPRR1308. Defaulting the brochure into the obligation is how a public brief becomes a false compliance conclusion.

graph TB
  Q["Queue request"] --> S["Study"]
  S --> C["Classification"]
  C --> A["Approval to energize"]
  A --> E["Observed energized"]

  P["Advertised campus MW"] -.->|"does not substitute"| D["Expected aggregate peak Demand"]

  style Q fill:#e8e8e8,color:#0a0a0a,stroke:#0a0a0a
  style E fill:#e8e8e8,color:#0a0a0a,stroke:#0a0a0a
  style P fill:none,stroke:#0a0a0a
  style D fill:none,stroke:#0a0a0a
          

A serious screen keeps three values. Known. Unknown. Cannot determine. An obligation that needs an unknown input cannot be marked as applying. It stays undetermined until the project owner confirms the missing record.

What survives the next hard question

The useful artifact is not another 466 GW dashboard. It is a project record that states what applies, what is known, what is missing, who owns the next action, and which source was current when the statement was made. Every rule claim needs a primary-source URL and a verified or unverified flag. Proposed text stays proposed. Unknowns stay unknown.

ERCOT's Large Load Integration page and dated market notices remain the process sources. Pallara's tracker at lli-radar.pallara.xyz mirrors those files. It does not replace them. A named-project screen from public facts lives at pallara.xyz/ercot.

This is decision support, not legal advice. It does not replace counsel, project engineers, or authorized representatives. Re-verify every deadline, fee, and megawatt figure against the cited source before relying on it.