A once-reliable economic-development pitch is running into tougher questions about who pays for the infrastructure behind AI. Texas, Pennsylvania and New York show how the political terms for data-center growth are changing.
Texas Gov. Greg Abbott halted approvals for some 1,800 new data centers, a decision that puts Texas at the center of a wider reversal in political support for the facilities powering the AI boom. The shift toward criticism and restrictions is being driven by concerns about electricity and water use, even as leaders continue to pursue major technology investment.
Abbott had celebrated a $40 billion Google investment in November, calling Texas the “epicenter of AI development.” Less than a year later, he said approvals for proposed facilities worth billions had been halted over their power and water needs, according to the Wall Street Journal.
Texas changes the AI equation
The significance of Abbott’s move lies in the contrast. Texas has aggressively sought technology investment, and data centers have been sold as sources of construction work, corporate spending and a prominent role in the expanding technology economy.
That case has not disappeared. President Donald Trump has defended data centers as an economic engine, reflecting the view that the facilities can help the United States compete for AI infrastructure and attract investment.
But a giant computing campus also requires constant, dependable electricity. Residents and policymakers are confronting whether new demand could require costly grid upgrades, affect reliability or add pressure to utility bills. Water required for cooling is another concern, particularly in communities watching their available supplies.
Texas carries particular weight in this argument because its electric system has faced close scrutiny over reliability. A rapid increase in power-hungry facilities creates a basic policy problem: how quickly can the state add major new demand without creating unacceptable infrastructure strain?
What Abbott’s halt leaves unanswered
Abbott said he halted approvals for about 1,800 proposed data centers because of worries about power and water needs. The report did not identify the affected projects, say how long the pause would last or explain the standards that will determine whether projects can advance.
Those details will determine whether the halt becomes a short-term reset or a lasting change in Texas’ approach. A pause could give officials time to establish clearer requirements. It could also mean a much higher bar for projects that expected state and local approvals to be more routine.
The decision is not necessarily a rejection of AI investment. It does, however, show that a large announced investment no longer guarantees an easy path for every related development proposal.
Pennsylvania chooses conditions over speed
Pennsylvania Gov. Josh Shapiro has taken a different approach: allowing the possibility of data-center development while attaching specific conditions to it. Executive Order 2026-05 requires proposed data centers seeking state permits to meet the administration’s Responsible Infrastructure Development, or GRID, requirements.
According to the state, developers must make legally binding commitments involving energy affordability, community engagement, workforce and economic development, transparency and environmental protection. The Pennsylvania Department of Environmental Protection is also directed to give greater weight to local approvals when evaluating applications.
The order removes AI data-center projects from the state’s Permit Fast Track Program and says they will not be eligible for that program going forward. Pennsylvania has also said nondisclosure agreements are not permissible for data-center projects, addressing demands from residents for clearer information before approvals are issued.
Shapiro framed the rules as a test of corporate responsibility: companies that want to build must meet strict requirements and win support in the communities where they plan to locate. Supporters may see that as a way to protect residents without walking away from investment. Developers may see added process, delay and uncertainty.
The fight is local, not abstract
Arguments about AI infrastructure often begin with national competition and billion-dollar announcements. The practical conflicts tend to emerge at the local level, where residents examine a particular site and ask about traffic, noise, land values, water availability, electricity prices and grid reliability.
Communities may be offered tax revenue, construction jobs or infrastructure improvements. Yet local officials and residents can still question whether those benefits justify tax incentives, land-use changes and the long-term utility demands associated with an always-on facility.
That is why the debate is more complicated than a simple choice between supporting or opposing data centers. Critics are often seeking enforceable environmental commitments, meaningful public notice, transparent utility planning and assurances that local ratepayers will not subsidize private computing demand.
Industry advocates offer a competing concern. They argue that slowing projects could send investment, technical talent and related business activity to other states or countries. They also argue that predictable rules are preferable to political changes after companies have begun planning expensive projects.
New York faces the same tradeoff
New York Gov. Kathy Hochul faces a related challenge: balancing technology and economic goals with electricity-system constraints and environmental commitments. The central question is not neatly partisan. It is how a state can welcome high-value digital infrastructure without transferring its costs to residents.
The answer may look different from one region to another. A project near abundant generation, strong transmission and adequate water can present a different case from a proposed facility in an area where the grid is already constrained.
That regional reality makes broad promises difficult to sustain. A pledge to approve every project quickly can collide with utility and water limits, while a blanket refusal can overlook locations with capacity and potentially valuable investment opportunities.
The new terms for data centers
The political shift does not mean AI development is ending. It means politicians are increasingly being pressed to define the public terms under which its infrastructure will be built.
For developers, a large investment figure may no longer be enough to secure support. Plans increasingly need credible answers on electricity procurement, water management, local engagement, workforce benefits and transparent permitting.
For elected officials, the difficult task is avoiding two outcomes at once: allowing projects to create unplanned local costs, while making the rules so uncertain that viable investment cannot move forward. Abbott’s Texas halt and Shapiro’s Pennsylvania guardrails suggest the next phase of the data-center boom will be decided less by enthusiasm for AI than by the conditions communities demand before hosting it.
Leave a Reply