Vivold Consulting
Policy & Regulation

Texas slams the brakes on data centres - and the AI buildout's easiest frontier just closed

Abbott orders audits of every new project as ERCOT's queue hits 474GW, roughly 90% of it data centres

Key Insights

Governor Greg Abbott announced that all new Texas data-centre projects must be audited by the Public Utility Commission and grid operator ERCOT - a sharp turn for a state whose loose regulation and cheap power made it second only to Virginia for data centres. The trigger is a staggering queue: ERCOT's interconnection requests doubled from 233GW in January to 474GW, about 90% data centres, more than five times the grid's all-time peak demand. Audits will demand power and water use, noise mitigation, light controls, tax-incentive use, and ownership details - after a voluntary survey that most operators simply ignored.

Stay Updated

Get the latest insights delivered to your inbox

Even Texas has a limit

Tech companies scouring the US for data-centre sites have been drawn to Texas by loose regulation and seemingly abundant power - only Virginia hosts more. That era may be ending. Governor Greg Abbott announced that all new data-centre projects will need to be audited by both the Public Utility Commission of Texas and the state's grid operator, ERCOT.

The number that forced the decision

ERCOT's interconnection queue has exploded. In January it held 233 gigawatts of projects waiting to connect; in under six months that more than doubled, and Abbott's office now puts it at 474 gigawatts of new connection requests - roughly 90% of them data centres, per the grid operator. For scale, that queue is more than five times ERCOT's all-time peak demand. Many entries are paper proposals, since long queues push developers to get in line early and plenty of projects fizzle - but even a fraction materialising could overwhelm the grid. Electricity prices in Texas remain affordable relative to other states, helped by utility-scale solar quadrupling between 2021 and 2025, but they have been rising, with the EIA attributing part of that to data centres and crypto mining.

What the audits will demand

Abbott has directed PUCT and ERCOT to collect on- and off-site electricity and water demand, noise mitigation, light controls, use of tax incentives, and ownership details. The heavier hand followed a failed lighter one: an earlier voluntary survey went mostly unanswered. In a state where Houston famously has no zoning code, that is a notable reversal - and TechCrunch's closing line is worth quoting in a board deck: depending on what the audits find, Texas's days as a data-centre mecca may be ending.

The infrastructure planning consequences

  • If your AI roadmap assumes cheap, fast US capacity, rebuild the assumption. The most permissive large market just added a review gate, and public opposition to AI infrastructure is rising nationally. Expect longer lead times and higher delivered costs on new capacity from 2027 onward.
  • For anyone signing multi-year cloud or colocation contracts, ask providers directly where contracted capacity will be sited and what its interconnection status is. A queue position is not a data centre, and providers with energised sites now hold real pricing power.
  • Water and noise disclosure is becoming standard. If you operate or sponsor facilities, get ahead of it: the operators who ignored Texas's voluntary survey converted a soft ask into a mandatory audit, which is a lesson in how regulatory escalation actually works.
  • Strategically, this strengthens the case for the efficiency lane: smaller models, better routing, open weights run locally, and inference optimisation all reduce exposure to a capacity market that is getting tighter and more politically contested. Cost engineering is now also risk management.

More in Policy & Regulation

All Policy & Regulation stories

Apple sues OpenAI for trade-secret theft - alleging the scheme ran 'at every level'

Apple sued OpenAI in federal court in Northern California for trade secret theft and breach of contract, alleging former Apple employees took confidential material to benefit OpenAI's consumer hardware ambitions - and that the misconduct was directed by senior leadership, running, in Apple's words, from members of technical staff to the Chief Hardware Officer. Specific claims include an engineer who allegedly kept an Apple laptop and downloaded confidential documents, and OpenAI allegedly using Apple's proprietary metal-finishing technique by misleading a shared supplier into believing it had permission. IO Products is also named. OpenAI says it has no interest in others' trade secrets.

AWS just published the ROI case for GraphRAG: drug research cycles cut by 87%

An AWS GraphRAG deployment in pharmaceutical research cut R&D cycles by 87% - initial discovery that took six months now closes in three weeks - by fusing siloed internal databases and public literature into one queryable knowledge graph on Amazon Neptune Analytics and Bedrock (running Claude). Every answer comes with verifiable citations and a mapped reasoning path, which is exactly what regulated industries need for compliance. The architecture is modular and, crucially, transferable: any enterprise drowning in fragmented legacy data can copy this pattern.

Alibaba blacklists Claude Code after hidden China-detection code surfaces - the AI decoupling gets personal

Alibaba has banned employees from using Anthropic's Claude Code from July 10, classifying it as high-risk software and directing staff to its in-house Qoder tool instead - some reports say the directive extends to uninstalling all Anthropic products. The trigger: a June 30 Reddit reverse-engineering post revealed the tool had quietly shipped code since April that checked for signs a user was in China (timezones, proxies, AI-lab infrastructure); an Anthropic engineer called it an anti-reseller, anti-distillation experiment that has since been removed. The backdrop is an open feud - Anthropic accused operators tied to Alibaba's Qwen lab of running a 28.8-million-exchange distillation attack via ~25,000 fraudulent accounts.