In the span of just six weeks, three major US states have successively put the brakes on AI data center expansion. On July 14, New York Governor Kathy Hochul signed an executive order imposing a one-year statewide moratorium on hyperscale data centers of 50 megawatts or more, making New York the first state in the nation to halt hyperscale computing infrastructure at the state level. On August 3, Texas Governor Greg Abbott ordered the Public Utility Commission of Texas (PUCT) and the Electric Reliability Council of Texas (ERCOT) to conduct a comprehensive review of all data center projects requesting grid interconnection, with projects barred from moving forward during the review period. On August 18, Pennsylvania Governor Josh Shapiro signed an executive order removing large data center projects from the fast-track approval pathway, requiring developers to make legally binding commitments — meeting the "Responsible Infrastructure Development" (GRID) requirements and obtaining local approvals — before the state Department of Environmental Protection would commence permit review. Additionally, according to a disclosure by the law firm Foley & Lardner on August 12, Chicago Mayor Brandon Johnson signed an executive order on August 11 directing the City Council to enact a temporary ban on new data center construction and large-scale expansions of existing facilities.
The simultaneous occurrence of these four actions is no coincidence. The core contradiction driving this trend can be seen clearly in a set of numbers: according to documents published by Texas utility regulators, ERCOT is currently processing approximately 474 gigawatts of grid interconnection requests — more than five times the state's historical peak electricity load — and roughly 90 percent of the new demand applications come from data centers. On this metric alone, the backlog of interconnection requests in Texas is no longer a mere queueing problem; it means that even if power supply capacity several times the historical peak were fully deployed, it would still fall far short of accommodating the scale of data center applications. Texas is not an isolated case. Pennsylvania's Department of Environmental Protection had recorded more than 100 data center proposals before Shapiro signed his executive order, of which 58 had formally initiated permit discussions with the DEP, 15 had submitted at least one permit application, and five had completed all approvals in the initial development phase.
Behind these numbers lies the structural tension that has been building at an accelerating pace in the AI computing race in recent years. According to an August 4 report by Reuters, Microsoft, Meta, Oracle, Amazon, and Alphabet — Google's parent company — have a combined approximately $1.09 trillion in lease payment commitments not yet fulfilled, most of which are tied to data centers required for AI. Google has previously announced plans to invest approximately $40 billion in Texas to build three new data center campuses, and the "Stargate" project — involving OpenAI, Oracle, and others — also designates Texas as a key base. Commitments of this magnitude signify one thing: the AI computing race has fully extended from the soft competition of chips and models to the hard contest over land, facilities, and power infrastructure.
The problem is that the electricity and water demanded by data centers are not consumed in isolation by the companies themselves. New generating capacity requires grid expansion, and the cost of upgrading transmission lines, absent a clearly designated party responsible for bearing it, tends to be passed on to surrounding residents and businesses through higher electricity rates. This is precisely the root of community backlash, and it is the policy logic behind the successive state actions.
The SPACE City survey released in May of this year by the Hobby School of Public Affairs at the University of Houston captures this structural contradiction in public opinion data: 85 percent of surveyed residents in the Greater Houston area said they use AI, yet 62.9 percent opposed building data centers within one mile of their homes. Among those who objected, 80 percent cited the enormous electricity demand as their primary concern. Across party lines, more than 75 percent of respondents believed that data centers and technology companies should bear the costs of related energy and infrastructure upgrades. Notably, public attitudes toward data centers are not monolithic — 32 percent of opponents said they might shift to support if the facilities ran on renewable energy rather than fossil fuels. This suggests that the heart of the controversy is not "whether we want AI" but "who pays the bill and who bears the risk."
The three states' choices of policy tools each have their own emphasis and their own boundaries of enforceability. New York's moratorium is the most direct: it freezes environmental permits for projects above 50 megawatts by executive order, and is accompanied by a plan to repeal the sales tax exemption for large data centers — if realized, this would directly raise developers' capital expenditures on key hardware such as GPUs and electrical equipment. But a moratorium is ultimately time-limited; whether it can evolve into a durable mechanism after the one-year term expires remains uncertain. Texas's review order sets up a checkpoint at the grid level: until PUCT and ERCOT complete their comprehensive review, new projects cannot advance the interconnection process — a substantial obstacle for data centers that depend on large volumes of cheap electricity. Yet whether the review can actually constrain large project access once completed depends on the final design of the review criteria. Pennsylvania's GRID framework imposes four categories of specific requirements on developers: first, assume all new electricity costs — generation, transmission, and distribution — without passing them on to residents and businesses; second, conduct open and transparent community engagement, including public meetings and early-stage design consultations; third, hire local workers and sign community benefits agreements; and fourth, meet the highest environmental standards, including stringent water conservation requirements. All of these requirements must be implemented in the form of legally binding commitments, setting a significantly higher bar than the previous fast-track pathway.
The impact of this policy shift is not evenly distributed across the various players in the AI infrastructure supply chain. For hyperscale cloud providers and AI companies, the most immediate effect is that site-selection strategies need to be recalibrated: traditional data center hubs where permits could previously be advanced quickly now carry added political risk and compliance costs. Foley & Lardner noted in a July 15 report that multiple states — including Minnesota, Michigan, South Carolina, New Hampshire, and Virginia — are considering similar legislation. If more states follow suit, this would fundamentally change the site-selection logic for new projects, with regulatory risk joining power supply, fiber connectivity, and tax incentives as a core evaluation dimension in infrastructure planning. For data center developers, project timeline uncertainty has intensified, and capital expenditure structures will also shift: the early phase now requires simultaneously addressing new generating capacity, grid expansion, energy storage, and water resources, rather than building first and connecting to the grid later. This means longer development sequences, front-loaded capital pressure, and higher negotiation costs with local governments and communities. As for ordinary residents worried about rising electricity bills amid the influx of data centers, these policies represent their interests to a certain extent — but whether the policies can truly bind, or remain at the level of executive orders lacking enforcement mechanisms, remains to be seen.
Set against the state-level restrictions is a parallel federal effort operating through a voluntary commitment mechanism, and the two share an underlying policy logic. On March 4, 2026, the White House brokered the signing of the "Ratepayer Protection Pledge" by major technology companies including Amazon, Google, Meta, Microsoft, OpenAI, and Oracle. The core provision: participating companies commit to self-building, procuring, or purchasing power resources to meet the incremental demand from data centers, and to bear the costs of upgrading new electricity transmission infrastructure, avoiding the pass-through of related costs to ordinary households. On July 23, the White House announced the pledge's expansion, with more than 200 additional utilities, data center developers, cooperatives, and states joining — claiming coverage of approximately 80 percent of US residential and commercial electricity consumption and protection for roughly 263 million Americans. This federal mechanism is highly consistent with Pennsylvania's GRID requirements in cost-bearing logic: both place the "AI expansion bill" back on technology companies rather than passing it to users. The difference is that the White House pledge is voluntary in nature, while the state executive orders carry legal binding force. Whether the two can form an effective combined force depends on the actual enforcement of the federal pledge and the judicial robustness of the state executive orders.
Viewed from a timeline perspective, this regulatory shift is occurring at the same time that AI infrastructure investment has entered a historic peak. According to Reuters, the outstanding data center-related lease commitments of leading technology companies total more than a trillion dollars. The gap between capital and regulation is narrowing simultaneously across multiple states.
Based on the above analysis, the following assessments are forward-looking projections rather than confirmed facts. The current policy trajectory is most likely to evolve along one of two paths. The first: the federal and state "cost internalization" mechanisms gradually harden into standards, and AI infrastructure construction enters a new phase with longer approval cycles, higher compliance costs, but more stable overall rules — similar to the "local backlash → standardized regulation" cycle experienced by US energy infrastructure expansion during the shale era. The second: state policies are substantially weakened under administrative and judicial review and industry lobbying pressure, some executive orders become mere formalities, and companies circumvent costs by relocating projects to regulatory-friendly states, producing a "regulatory arbitrage" landscape. Two key signals will determine which path is more likely. First: whether wait-and-see states such as Minnesota and Michigan enact legislation rather than stopping at executive orders before the end of 2026 — legislation would be far harder to overturn through a single administrative mechanism. Second: whether Pennsylvania's GRID framework produces the first batch of rejected permit applications — only real rejection cases can verify whether the new rules have actual binding force, rather than serving as political signals alone.
Sources: - [Governor Shapiro Signs Executive Order on Data Center Development in PA](https://www.pa.gov/governor/newsroom/2026-press-releases/governor-shapiro-signs-executive-order-on-data-center-developmen) - [Facing an estimated 474 GW of interconnection requests, Texas hits pause on data centers | Utility Dive](https://www.utilitydive.com/news/texas-hits-pause-data-center-interconnections/827046/) - [New York's AI Data Center Moratorium: A Potential Turning Point for U.S. Data Center Development | Foley & Lardner](https://www.foley.com/insights/publications/2026/07/new-yorks-ai-data-center-moratorium-a-potential-turning-point-for-u-s-data-center-development/) - [First Statewide Moratorium on New Hyperscale Data Centers Launched by Governor Kathy Hochul](https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul) - [Majority of Houstonians use AI but oppose data centers near their homes, Hobby School survey finds](https://www.houstonpublicmedia.org/articles/news/energy-environment/2026/05/19/552380/majority-of-houstonians-use-ai-but-oppose-data-centers-near-their-homes-hobby-school-survey-finds) - [President Trump's Ratepayer Protection Pledge Secures American AI Dominance, Protects Consumers](https://www.whitehouse.gov/releases/2026/07/president-trumps-ratepayer-protection-pledge-secures-american-ai-dominance-protects-consumers/) - [Abbott Orders Audit of Texas Data Centers Seeking Grid Access](https://news.bgov.com/bloomberg-government-news/abbott-orders-audit-of-texas-data-centers-seeking-grid-access)© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接