Data centres for AI and cloud computing need large amounts of land, electricity and water. In parts of Texas, local Republican officials and voters have begun resisting new projects, complicating an industry that had expected an easy welcome.
Key takeaways
- Opposition to new data centres in Texas has emerged from within Republican-leaning communities rather than from the environmental groups usually associated with infrastructure objections.
- The core local concerns are electricity demand, water use for cooling, noise from cooling equipment and the effect of large industrial sites on rural land and property values.
- Texas has attracted data centre investment because of its deregulated electricity market, available land and relatively permissive permitting, which is now producing concentrated local impacts.
- Technology companies expanding artificial intelligence capacity face schedule and siting risk when local approvals, utility interconnection and tax incentives become contested.
- The dispute is largely a local-government matter, so outcomes will vary between counties and municipalities rather than being settled in a single decision.
What is actually happening
Across parts of Texas, proposals to build large data centres — the warehouse-scale buildings that house the servers behind cloud services and artificial intelligence systems — have run into local resistance. What makes this notable is the political composition of that resistance. Much of it comes from residents and elected officials in conservative, Republican-leaning areas that have historically been receptive to industrial development and to arguments about jobs and tax base.
The objections raised in such disputes typically fall into a small number of categories: the amount of electricity a facility will draw from the regional grid, the volume of water used for cooling, the noise generated by cooling systems running continuously, the visual and traffic effect of large industrial buildings in rural or semi-rural settings, and the value of tax abatements offered to secure the investment relative to the small number of permanent jobs a data centre creates once built.
At the same time, technology companies are in an unusually intense phase of building computing capacity for artificial intelligence. That combination — accelerating demand for sites and hardening local resistance in a state that had been among the most accommodating — is the substance of the story.
Why this is in the news now
Two trends have converged. The first is the scale and pace of AI-related construction. Training and running large models requires far more computing hardware, and therefore far more power per building, than the previous generation of cloud facilities. Sites that once drew tens of megawatts are now proposed at much larger capacities, which changes how visible they are to a local grid and a local community.
The second is that the effects of earlier construction have started to be felt. Once a first cluster of facilities is built in an area, subsequent proposals are judged against lived experience rather than promises: residents can point to what changed after the last one. Local politics tends to respond to that shift.
Precisely how many projects have been delayed, refused or withdrawn, and in which jurisdictions, is not something that can be stated reliably here. Local planning decisions are recorded county by county and city by city, and there is no single authoritative tally. Readers should treat any specific figure they encounter with caution unless it names the jurisdiction and the decision.
The background a newcomer needs
Texas has been attractive to data centre developers for structural reasons. It operates its own electricity market, largely separate from the two big interconnections that cover most of the rest of the United States, which has historically meant relatively fast connection processes and competitive wholesale prices. Land is comparatively cheap and plentiful. State and local governments have offered tax incentives for large capital investments. Zoning rules in unincorporated areas are often lighter than in other states.
Those same features create the friction now visible. A grid that connects new load quickly also exposes existing customers sooner to the consequences of that load. Cheap rural land places industrial buildings next to homes that were not previously near industry. Tax abatements that attract investment also reduce, at least temporarily, the local revenue that residents expect development to generate.
It is worth being precise about what a data centre is and is not. It is a building full of computing and networking equipment, supported by power distribution, backup generation and cooling. It employs a large workforce during construction and a much smaller one thereafter — typically technicians, security and facilities staff. Its main ongoing local footprint is electricity, water or air used for cooling, and noise.
Who is affected, and how
Residents living near proposed sites are affected most directly, through noise, construction traffic and changes to the character of the land around them. Electricity customers across the region are affected indirectly, because large new loads influence the balance of supply and demand and the need for new transmission and generation, the costs of which are ultimately recovered through rates. How those costs are allocated between the new load and existing customers is a central and genuinely contested question.
Local governments face a trade-off between capital investment on the tax roll and the political cost of approving unpopular projects. Utilities and grid operators face planning problems, because a proposed facility may or may not be built, and developers sometimes seek interconnection at several sites for a project that will only be built at one.
Technology companies and the developers who build for them face schedule risk. Capacity plans assume sites can be secured and energised on a timetable; contested approvals introduce uncertainty into that timetable. Construction workers and contractors gain in the short term wherever projects proceed.
Where informed people disagree
There is real disagreement, not merely partisan noise, on several points.
One is cost allocation. Some argue that large loads should bear the full incremental cost of the generation and transmission they require, through dedicated tariffs or by bringing their own power. Others argue that large, predictable loads help spread fixed network costs across more consumption and can, on balance, benefit other customers.
A second is the value of the incentives. Supporters point to substantial capital investment, property tax revenue after abatements expire, and construction employment. Critics point to the small permanent workforce and argue the same public money could support development that employs more people.
A third is water. Cooling requirements vary considerably with design: some facilities use substantial water, others use closed-loop or air-cooled systems that use very little but consume more electricity. General claims about data centre water use are often applied to facilities whose designs differ materially, and comparisons are only meaningful when the design and the local water source are specified.
A fourth is whether current demand forecasts will hold. If AI-related demand grows as projected, capacity constraints are the binding problem. If projections prove too high, some announced projects will not be built, and infrastructure procured on their behalf may be underused.
The practical implications
The immediate practical effect is that siting has become a bottleneck alongside chip supply and electrical equipment. Developers increasingly need to secure local political consent, not merely land and a grid connection, and that requires earlier engagement, more detailed commitments on noise and water, and sometimes payments or agreements with local authorities in place of standard abatements.
For state policymakers, the questions concern how large loads are treated in market rules: whether they can be curtailed during scarcity, what they pay for interconnection, and whether they must supply some of their own generation. Rules of this kind are technical but determine much of the economic outcome.
For residents, the practical implication is that the decisions that matter are usually local and procedural — zoning changes, special use permits, abatement agreements — and are made in meetings that are open but not widely attended.
What to watch next
Watch whether resistance broadens from individual project fights into changes in state-level rules governing how large electrical loads connect and pay. Watch how tax incentive programmes are renewed or altered, since these are set by legislatures and have a clearer paper trail than local decisions. Watch whether developers shift towards sites with on-site generation, or towards other states and countries.
Watch, too, for cooling designs that reduce water use, and for agreements in which facilities commit to reducing consumption during grid stress. Finally, watch whether AI capacity demand continues at its current pace, because the strength of that demand determines how much developers are willing to concede locally to obtain a site.
Frequently asked questions
Why do data centres use so much electricity?
A data centre houses thousands of servers that run continuously, plus the cooling systems needed to remove the heat they generate. Facilities built for artificial intelligence use specialised processors that draw considerably more power per unit than conventional servers, so a modern AI-oriented site can require as much electricity as a small city’s worth of load, concentrated at a single connection point on the grid.
Do data centres create many permanent jobs?
Construction employs large numbers of workers, often for a year or more. Once operating, a data centre is largely automated and typically employs a comparatively small permanent staff of technicians, engineers, security and facilities personnel. This gap between construction employment and operating employment is one of the main points of contention when tax incentives are debated by local officials.
Why has opposition appeared in conservative areas?
The objections raised are largely about property, land use, local control and household electricity bills, which are concerns that do not map neatly onto conventional environmental politics. Rural and exurban residents near proposed sites experience noise, traffic and landscape change directly, and scepticism about tax abatements for very large companies exists across the political spectrum.
How much water do data centres actually use?
It depends on the cooling design. Evaporative cooling systems use significant quantities of water but less electricity; closed-loop and air-cooled designs use very little water but more power. Because designs vary so much, general figures are often misleading. Meaningful assessment requires knowing the specific facility’s cooling technology and the local water source it would draw from.
Can local governments actually stop a data centre?
Local authority varies by jurisdiction and by what the project needs. Where a project requires rezoning, a special use permit or a negotiated tax agreement, local bodies have substantial leverage. Where land is already appropriately zoned and no incentive is sought, that leverage is much weaker. Utility interconnection processes are usually governed at state or regional level instead.
Will this stop AI data centre construction in Texas?
There is no basis for predicting that. Texas retains structural advantages in land, energy resources and market rules, and local resistance is uneven across the state. The more likely effect is that projects concentrate in jurisdictions that remain welcoming, that developers negotiate more commitments to secure approval, and that some proposals move elsewhere.
Sources and further reading
- Regional electricity grid operator and state utility commission materials, which publish load forecasts, interconnection queues and market rule proceedings relevant to large new loads.
- Local government records — county commissioners’ court agendas, municipal planning commission minutes and tax abatement agreements — which are the primary documentation of individual siting decisions.
- Reporting by Texas state and regional news organisations, which cover local hearings and community responses in more detail than national outlets.
- Technology industry trade publications and company infrastructure disclosures, which describe capacity expansion plans, cooling technologies and site selection criteria.
Surfaced from the reddit:technology signal “local opposition to data centres”. AI-assisted draft, editorially reviewed.
