Texas Pauses Data Centers in a 474 GW Queue • The Weighted Average

Texas has paused data-center projects advancing through ERCOT until regulators complete a new five-part audit, confronting a queue of more than 474 gigawatts. The trigger is as revealing as the scale: only 28 of 377 recipients answered a state power-and-water survey, a derived response rate of just 7.4%.

A queue position now requires a proof package

Governor Greg Abbott’s official August 3 directive orders PUCT and ERCOT to complete the audit before any data-center project advances. Projects that fail the agencies’ requirements must be denied a grid connection. This is not merely another study of future electricity demand; it is a verification gate on project identity, public subsidy, resource dependence and local impact.

The state demands five categories. Developers must disclose tax incentives, grants, abatements and other public assistance; annual and peak power demand plus on-site generation; annual and peak water use, supply and cooling technology; community protections covering noise, light, setbacks, traffic and emergency response; and ownership and controlling interests. That turns interconnection diligence into a joined financial, electrical, environmental and beneficial-ownership record.

The arithmetic explains the skepticism. Abbott says ERCOT is considering more than 474 GW of connection requests and approximately 90% are data centers. Multiplying those approximate inputs gives roughly 427 GW of data-center requests. It is a queue estimate, not a forecast of simultaneous consumption or completed construction, but it is too large to evaluate as if every form represented an equally mature project.

The Texas Tribune reports more than 1,800 projects in the queue and says ERCOT put its Batch Zero transmission-planning study on pause after the order. The Tribune has identified at least 335 operating data centers and 248 planned facilities in Texas, while noting that on-site generation and projects outside ERCOT may avoid this particular grid path. Scope and duration remain unclear, which makes schedule contingency essential.

The pause lands on top of a formal process rather than an empty queue. ERCOT’s Large Load Integration page requires current forms, utility validation and an approved energization request for qualifying projects. That process can screen engineering readiness; the new directive adds project-wide verification before advancement.

The survey response is the sharper project-quality signal. PUCT staff said it notified 377 companies, while 28 responded. Dividing 28 by 377 yields 7.4%. A response is not proof that a project is viable, and silence may include stale contacts or abandoned proposals, but such a low participation rate gives the state little basis for long-term water and power planning.

This audit is distinct from yesterday’s NERC equipment-behavior requirements. NERC asks how energized computational loads, UPS systems and protection controls behave during disturbances. Texas is asking whether proposed projects have credible incentives, power, water, community and ownership records before they move forward. One governs dynamic behavior; the other tests whether the project exists in an approvable form.

It also changes the economics behind SpaceX’s rapid AI-capacity buildout. Servers and contracts cannot earn revenue until a site clears political as well as electrical diligence. The audit makes an incomplete evidence room a direct threat to energization, not a paperwork inconvenience.

Freeze the schedule that cannot survive disclosure

Developers above the large-load threshold should audit themselves before Texas does. Assemble executed or conditional incentive agreements, site control, a complete ownership chart, annual and peak electricity and water models, cooling design, on-site generation status, permits, traffic and noise studies, emergency plans and a phased ramp. Reconcile every figure across ERCOT, PUCT, municipal and investor submissions; inconsistencies will be easier to find when five workstreams meet in one review.

The cost begins with delay. The directive gives no completion date, and the Tribune says Batch Zero planning was postponed. A developer carrying land, equipment reservations and financing can lose meaningful value even if it eventually passes. Procurement should therefore condition long-lead accelerator and electrical orders on audit milestones, while financing models add a regulatory-delay case rather than assuming a queue date equals an energization date.

The underlying statutory screen already imposes a floor. Texas SB 6 sets a 75 MW default threshold, at least a $100,000 initial fee and disclosure of backup generation capable of serving 50% of demand. Those are enacted requirements, unlike several still-proposed tariff and collateral figures. The audit now asks whether the whole project story supports the electrical application.

Grid-connected projects should also model alternatives without pretending they are free. On-site generation may reduce ERCOT dependence, but it adds fuel, emissions, permitting, redundancy and maintenance questions. Air cooling can reduce water demand but may increase energy use or constrain density in extreme heat. Closed-loop systems change replenishment rather than abolish water risk. The audit correctly forces these trade-offs into the same commercial plan.

This is a stronger requirement than the older queue process but not evidence that Texas has rejected data centers. The state can use verification to remove paper projects and accelerate credible ones. The Data Center Coalition told the Tribune that a well-run review could showcase responsible developers rather than delay them. A low 7.4% survey response rate supports triage; it does not prove the remaining 92.6% are irresponsible.

There are also legal and implementation risks. The audit’s scope is not yet precise, PUCT had not described a completion timetable, and agencies must integrate it with an already changing interconnection process. A blanket pause can strand mature projects alongside speculative ones. Ownership disclosure, local tax terms and infrastructure plans may also involve confidential business information that needs a secure, consistent submission channel.

What would change the verdict? Confidence rises if PUCT and ERCOT publish a uniform checklist, confidentiality rules, project-level deficiency notices, a restart sequence and service-level deadlines—and if mature projects clear without serial information requests. Confidence falls if the pause has no timetable, criteria change mid-review or agencies cannot distinguish a financed site from a placeholder request.

For operators, the decision is immediate: pause schedules that lack a five-part proof package, not every Texas project. Developers with complete records should use the audit to separate themselves from speculative capacity. Teams comparing markets should price the verification delay beside land, tax and power costs; our earlier look at the Nvidia-Hut 8 Texas lease showed why a nominally attractive capacity agreement still depends on the site beneath it.

Texas has converted “how many megawatts do you want?” into “show who owns the project, who subsidizes it, and what it consumes.” That is a higher bar, but for a queue larger than the grid it serves, paperwork was never the scarce resource. Credible projects were.