AI News
Texas paused data‑center approvals while regulators audit a 474‑gigawatt queue, shifting AI infrastructure risk back to developers and investors.
ERCOT is tracking roughly 474 gigawatts of new connection requests. The governor's office says about 90% are data centers, and the total is more than five times Texas's record peak electricity demand.1
Not every request will become a building. Developers often seek queue positions before financing, permits and customers are final. But an inflated queue still imposes costs: grid planners cannot treat hundreds of speculative projects as either real or irrelevant. The audit is an attempt to separate credible capacity plans from paper options.
The review reaches beyond megawatts. Applicants must provide projected electricity demand, on‑site generation plans, water use, cooling technology, tax incentives, ownership and community protections such as noise controls and setbacks.1
That turns infrastructure diligence into a whole‑project test. A developer can no longer sell “AI capacity” while leaving ratepayers, municipalities and grid operators to discover the water and transmission consequences later. The state is asking who pays, who owns the project and what happens outside the server hall.
Texas remains attractive because of natural‑gas supply, renewable generation and a business‑friendly regulatory environment. Yet data centers are unusually concentrated loads: they can require the electricity of a city while arriving faster than new transmission or generation.2
The investor consequence is simple. A site with low nominal power prices is not cheap if interconnection timing is uncertain, new generation is required or the developer must fund grid upgrades. The constraint shifts from buying chips to securing credible electricity, water and political permission.
The state has not announced a fixed end date or prohibited all future data centers. Projects can proceed after the PUCT and ERCOT complete their verification and determine that requirements are met.1
That caveat matters for headlines and financial forecasts. “Texas bans AI” would be wrong. The more accurate conclusion is that grid access has become conditional on evidence—and the timeline for satisfying that condition is not yet clear.