Texas wants to be the capital of the artificial-intelligence boom. But the rush to build the giant data centers powering that ambition has collided with a basic constraint: the grid cannot approve every project on the strength of a developer’s promise, and residents should not be left to discover later what the buildout will cost them.
That tension is now playing out in Texas. After the state’s grid operator, ERCOT, received an extraordinary volume of requests from large electricity users, Governor Greg Abbott ordered a review of proposed data-center projects seeking to connect to the grid. In September, he widened the scrutiny, directing state environmental regulators to pause certain permits for data-center infrastructure while state agencies assess power and water impacts. The pause is not a ban on data centers, and it does not mean every project in the queue would be built. It does mark a change in a state that has spent years advertising itself as a place where big projects can move quickly.
ERCOT large-load connection requests
The queue helps explain the concern—but it must be read carefully. ERCOT reported that large-load connection requests totaled about 239 gigawatts in February. Later estimates were even higher. These are applications and proposals, not a forecast of electricity that will definitely be needed: many projects may never secure financing, customers or permits. Still, ERCOT has said data centers and cryptocurrency operations make up a large share of anticipated demand, creating a difficult planning problem for a grid already managing rapid growth.
The question is not whether Texas should welcome investment. Data centers bring construction work, major private investment and a new tax base. They can also support local businesses and spur upgrades to infrastructure. The question is whether the state can capture those gains without making a risky bargain—offering public incentives and scarce resources before it knows what a project will demand, what it will deliver and who will pay for the power to run it.
A queue is not a forecast
The national investment figures show why the pressure is growing. Amazon, Alphabet, Meta and Microsoft are expected to spend a combined $725 billion on capital projects in 2026, based on estimates compiled from their first-quarter earnings. That figure covers more than data centers, but the companies’ expanding AI infrastructure is a major part of the investment surge. Meta alone has forecast capital spending of up to $145 billion this year.
The buildings and computing equipment can be planned and financed on corporate timelines. Power generation and transmission do not always move at the same speed. New substations and lines require studies, approvals, equipment and construction. In the PJM region, for example, a fast-track process approved three projects totaling 2.1 gigawatts of storage and gas-generation capacity, with completion expected by mid-2029. That is a substantial addition, but it is not an immediate solution to demand growing now.
Texas is trying to avoid approving large loads one at a time without seeing their combined effect. ERCOT is moving toward a batch process to assess proposed connections together. Abbott’s audit adds scrutiny of projects’ expected electricity and water use, on-site generation plans, tax incentives and community impacts.
There is a sound reason to check those details before approving new connections. But a large queue should not be mistaken for a list of projects certain to be built. The risk runs in both directions: overestimating demand could prompt expensive infrastructure that is not needed; underestimating it could leave the grid short of capacity when projects do arrive. Good planning requires distinguishing serious, financeable projects from speculative requests—and updating those assessments as projects change.
Growth has a price
The economic case for data centers is real, but it should be measured against the public cost of attracting them. Texas offers tax incentives to qualifying data centers, and projects can also seek local economic-development support. These incentives may help secure investment, but every tax exemption is revenue the state or local government forgoes—money that could otherwise support schools, public services or infrastructure. A project’s headline investment figure is not, by itself, proof that the public receives a comparable return.
Nor is job creation a simple calculation. Construction can generate substantial work while a facility is being built; the number of permanent jobs once it is operating may be much smaller. If governments are offering incentives, residents should be able to see the expected jobs, the public subsidy and the conditions attached to it—and later compare the promise with the result.
The rapid expansion of AI data centers across Texas is part of a broader U.S. construction boom that is reshaping how technology companies build computing infrastructure. Microsoft’s revised plans for its Mount Pleasant, Wisconsin, campus reflect the same trend, as surging demand for AI computing and cloud services drives developers to expand capacity and rethink facility designs. From Texas mega campuses to Wisconsin developments, the race to support AI is fueling large-scale construction, increasing demand for electricity and putting data center projects at the center of local planning and infrastructure decisions.
Electricity costs
Electricity costs are more complicated still. The argument is not that every data center automatically raises every household’s bill. Large customers can pay for dedicated infrastructure and contribute to a utility’s fixed costs. Grid improvements may also benefit other customers, strengthen reliability or serve future demand beyond one facility. But when the scale or timing of a project drives new upgrades, regulators need a transparent method to decide what the company pays and what, if anything, should be shared.
A recent investigation by Democratic senators found that major data-center companies support paying for infrastructure that serves their facilities alone but dispute paying the full cost of broader grid upgrades their demand may trigger. That is the heart of the dispute: who pays for shared infrastructure when one customer’s arrival makes the investment urgent, but other customers may eventually benefit too?
The answer should not be decided behind closed doors. The senators’ investigation also found that companies sought tax incentives and, in some negotiations, nondisclosure agreements that limited what communities could learn. Confidentiality may help companies negotiate land or commercial arrangements, but it should not obscure the use of public funds or prevent residents from understanding a project’s likely demands on power, water and local infrastructure.
Texas’s test
Texas has begun to set conditions. The state has directed regulators to scrutinize the grid and water implications of data-center proposals and to consider how developers will cover infrastructure costs. That is a necessary starting point, not proof that the public is protected. The safeguards will matter only if regulators can verify the information, enforce the rules and explain their decisions.
That scrutiny should not become a blanket rejection of the industry. Texas has a genuine opportunity to attract investment, expand its tax base and modernize an electricity system that must serve more than data centers. Developers should be able to build when they can demonstrate that their projects are viable, their power and water plans are credible, and their financial commitments match the demands they place on public infrastructure.
The state should make those tests clear: require reliable disclosure of project-level electricity and water use; publish the public incentives and expected benefits; ensure developers fund infrastructure built specifically for them; and use an independent, transparent method to allocate costs for upgrades that serve both a project and the wider grid. If projects promise lasting jobs or lower bills, those claims should be measurable and checked after the incentives are awarded.
The AI buildout is moving faster than the grid, but speed alone is not a public benefit. Texas can welcome the investment without asking households and communities to accept a blank cheque. Before another data center gets the power to grow, the public should be able to see what it will consume, what it will contribute and who will be left paying if the promises fall short.
Leave a Reply