Summary

  • Governor Greg Abbott directed the Public Utility Commission of Texas and ERCOT on August 3 to verify and audit every data-centre project advancing through ERCOT’s interconnection process before any project moves forward.
  • The governor’s office says ERCOT is considering more than 474GW of connection requests, over five times Texas’s record peak demand, with approximately 90% of new power requests attributed to data centres.
  • The 474GW figure is requested capacity in a development queue; it is not built capacity, committed demand or evidence that the same load will ever operate simultaneously.
  • Projects are to disclose public financial assistance, annual and peak power requirements, self-generation plans, annual and peak water demand, water sources, cooling technology, community measures, location, ownership and controlling interests.
  • The official release says a project that fails applicable PUCT, ERCOT or state-law requirements must be denied connection, but it does not publish an audit deadline, methodology, project list or findings.
  • The directive does not close existing data centres, reject every queued project or prove that anticipated demand has already damaged reliability, rates or water supplies.

A connection queue is a claim on the future

An interconnection request is the opening balance of a negotiation, not a meter reading. Developers may apply before land, customers, financing, equipment and construction schedules are final. Several concepts can compete for the same site or commercial opportunity. A single project can change size as power becomes available. Some applications mature into operating load; others wait, shrink, move or disappear.

That is why the headline number needs a denominator and a stage label. More than 474GW is enormous beside ERCOT’s record peak demand, but the comparison does not mean Texas must serve 474GW tomorrow. It signals that the administrative queue has become too large to treat every requested megawatt as equally credible. If a regulator plans transmission or generation around speculative entries, it can overbuild. If it dismisses the queue as fiction, it can miss a genuine step-change in demand.

The audit introduces an intermediate task: determine which claims are mature enough to influence decisions. That means testing the relationship among requested capacity, project control, energy supply, water, local obligations and ownership. The aim is not to predict the future perfectly. It is to stop unlike projects from occupying the same undifferentiated line.

The state is moving the burden of proof toward applicants

The required information covers four resources that are often discussed separately. Electricity disclosures ask how much annual and peak demand a project expects, whether it will bring generation and what it can do to reduce pressure on ERCOT. Water disclosures ask the same questions about annual and peak use, supply and cooling method. Financial-assistance disclosures expose how much public support sits behind a nominally private investment. Ownership information identifies who ultimately controls the load request.

Together, those fields make the applicant explain its full operating proposition. A data-centre campus is not only a building seeking a connection. It is a bundle of capital, servers, cooling, grid upgrades, land rights, water arrangements and corporate control. A queue entry that supplies only a large MW number is therefore incomplete evidence.

The directive does not reveal the templates, verification standard or sanctions for an inaccurate submission. Nor does it say how PUCT and ERCOT will distinguish a well-developed project from an early option. Those implementation details will determine whether the audit becomes serious diligence or another questionnaire. The important institutional shift is clear, however: progression is made conditional on evidence rather than assumed from an application alone.

Peak demand, annual energy and flexibility must stay separate

A requested peak of 500MW describes a maximum rate of consumption. It does not reveal annual MWh, the hourly profile, ramp behaviour or whether part of the workload can move in time. Two campuses with the same peak can impose very different requirements on generation, transmission and reserves. A training cluster may sustain high load for long periods; an inference fleet may vary with user demand; maintenance and phased occupancy can change both.

The audit asks for annual and peak consumption, which is a useful minimum distinction. It should allow reviewers to test whether the project’s energy story matches its maximum connection request. Self-generation and demand-reduction measures add another layer, but they need operational definitions. Nameplate generation is not dependable capacity unless fuel, interconnection, availability and dispatch obligations support it. A promise to curtail is valuable only if triggers, duration, notice and verification are clear.

Nothing in the announcement establishes that any applicant has misrepresented these quantities. It also does not publish a common load-factor assumption. The audit can nevertheless reduce a basic category error: treating maximum requested power, expected annual energy and firm coincident demand as interchangeable numbers.

Water and cooling become connection evidence, not a footnote

Texas is asking projects to state projected annual and peak water consumption, the anticipated source and the cooling technology. That puts cooling architecture inside infrastructure diligence rather than leaving it for a later environmental debate. Air cooling, closed-loop systems and other designs create different trade-offs among water, electricity, capital cost and performance.

The requested disclosure does not automatically establish a water limit. It also does not say which source is acceptable, how drought conditions will be modelled or whether recycled water receives different treatment. A closed-loop label alone cannot answer every question: operators still need make-up water, heat-rejection equipment and contingency plans, with results shaped by climate and workload.

Even so, requiring a peak and annual account makes a project testable. Local utilities can compare the application with treatment, pipeline and drought capacity. Communities can distinguish a gross claim from a source-specific plan. Grid reviewers can see whether a cooling choice reduces water consumption by increasing electricity demand. The value lies in exposing linked constraints before a connection becomes difficult to unwind.

Public support can be measured against private readiness

The state wants details of tax incentives, grants, abatements and other public financial assistance already received or expected. That information matters because an interconnection request can create public planning work long before a facility produces tax revenue or employment. If benefits are contingent while network costs are immediate, the timing and allocation of risk deserve scrutiny.

Disclosure alone does not prove a subsidy is excessive. Local incentives may buy durable tax base, construction activity or strategic capacity; they can also transfer value to a project that never reaches operation. The audit could improve comparison by linking assistance to concrete stages such as site control, financing, energisation, occupancy and job delivery.

The announcement does not provide incentive amounts or a project-by-project ledger. It therefore cannot support conclusions about who is receiving aid or whether ratepayers carry a specific cost. Its significance is procedural: public support, grid dependency and project maturity are to be visible in one review rather than argued in separate forums with different numbers.

Ownership disclosure is infrastructure risk control

Data-centre capacity is frequently developed through special-purpose companies, joint ventures, lease structures and multiple contracting layers. The name on a connection request may not reveal the capital provider, eventual operator, customer or controlling interest. Requiring ownership and control information gives reviewers a better chance of understanding who can fund the project and who remains accountable if plans change.

This is not a licence to infer foreign control, security risk or financial weakness where none is disclosed. The official material names no suspect applicant. Ownership transparency is useful because obligations travel through entities: deposits, upgrade costs, curtailment commitments and decommissioning duties are only as meaningful as the party bound to them.

A mature audit should therefore connect corporate identity to enforceable responsibilities without confusing a developer, tenant, hyperscaler and financier. They may share an economic project while carrying different liabilities. The directive identifies the field; detailed PUCT and ERCOT practice will show whether it can follow those relationships through the interconnection lifecycle.

Denial is a boundary, not yet an outcome

The governor’s release says projects that fail applicable requirements must be denied connection. That statement gives the audit consequence, but it is not a list of denials. No project findings, rejection counts or completed reviews were published with the announcement. Existing operating facilities are not described as subject to shutdown.

This boundary matters for both critics and developers. Critics should not count all 474GW as approved growth. Developers should not treat an application position as a right to capacity. The decision surface lies between those extremes: evidence is checked, requirements are applied, and only then can a project progress.

Due process will matter. Applicants need to know what constitutes completeness, how inconsistencies are corrected, whether queue position is preserved during review and how decisions can be challenged. Grid planners need consistent data quickly enough to avoid replacing a congested queue with an audit backlog. The release is a mandate to build that process, not proof the process already exists.

Good queue governance should improve the signal

The audit’s success should be measured by the quality of the remaining queue, not merely by how many applications it removes. Useful indicators would include verified project control, secured funding stages, realistic energisation dates, transparent withdrawal rates, posted financial security and consistent definitions of peak, annual and flexible load. Publishing aggregates could improve planning without exposing commercial secrets.

A cleaner queue helps several parties. Utilities can plan upgrades against more credible demand. Generators and transmission developers can see where load is likely to materialise. Communities can assess water, noise and land effects before local promises harden. Developers with mature projects may face less competition from speculative applications for study resources and public attention.

There is also a risk that a broad audit becomes a political filter whose standards shift from project to project. Comparable fields, documented methods and explainable outcomes are the protection against that. The most durable result would not be a dramatic one-time purge. It would be a repeatable discipline that keeps the queue informative as applications enter, change and exit.

What the directive can and cannot establish today

The operative fact is that Texas has inserted verification before further data-centre advancement through ERCOT interconnection. The state has also defined a broad evidence set spanning electricity, water, public support, community effects and control. Those choices make the directive more than a warning about AI demand.

What remains unknown is equally important. There is no published audit completion date, no project roster, no conversion probability for the 474GW, no estimate of duplicated requests and no resulting grid plan. The directive does not prove that every request is speculative, or that every mature project will pass. It gives PUCT and ERCOT authority and responsibility to separate the two.

The Office of the Texas Governor’s announcement sets out the mandate and information fields. The directive letter is the primary instrument, while KXXV’s timestamped report records the public release in the window.

Sources

The three linked records above define the published evidence. Project-specific audit results, denials and implementation rules should be treated as future facts only when PUCT or ERCOT releases them.