Trump wants to massively build data centers; many places oppose it—will this become a new AI bottleneck?

TL;DR

· New York State suspends some newly built large data center state-level environmental permits, with a maximum term of one year.

· The controversy shifts to power, water resources, community costs, and AI infrastructure delivery timelines.

· Related tickers: ORCL, MSFT, GOOG, AMZN, Blackstone, as well as power, nuclear power, natural gas, liquid cooling, and data center developers.

On July 14, New York Governor Kathy Hochul signed an executive order to suspend state-level environmental permits for newly built hyperscale data centers for up to one year. It applies to large projects with electricity demand of 50MW or more. Trump later criticized that such restrictions would weaken the United States’ advantage in competing in AI and data centers.

For investors, the key is not whether New York will block all projects, but that the constraint conditions for AI infrastructure have changed. In the past, the market mainly focused on GPUs and capital expenditure budgets; now it also has to look at power, water resources, land, and local approvals.

Hyperscale data centers can be understood as massive power-hungry factories for AI. Project load could range from tens of megawatts to over a thousand megawatts, and will also pull in cooling water, grid interconnection, noise, and land use. The stronger the compute demand, the harder it is for these costs to remain solely on the company’s books.

Hochul’s logic is that New York cannot keep greenlighting giant projects without unified rules. Trump’s logic is the opposite: supporters argue that data centers represent investment, jobs, and compute advantages, and that the pause will push projects to other states. This conflict summarizes the contradictions in the next phase of AI infrastructure.

New York turns community backlash into an approval variable

The New York pause is important not because it will permanently ban data centers, but because it raises the friction—previously dispersed across resident hearings, town councils, and environmental lawsuits—up to the statewide approval level.

According to the governor’s official website wording, the executive order requires New York to advance a general environmental impact assessment and establish review standards centered on energy, water resources, and community impacts. In project terms, it means pressing the pause button first, then setting rules for how hyperscale data centers use electricity and water, compensate communities, and bear external costs.

The market can no longer simply extrapolate AI infrastructure with “demand is strong, so it will definitely get built.” Demand remains, but the construction path has narrowed. Whether a project can land depends on whether electricity prices are pushed higher, who pays for grid upgrades, whether regions with tight water resources can handle the strain, and whether tax incentives can win support from residents.

For tech companies, this kind of friction may not change the direction of long-term expansion, but it will change delivery timelines. In valuation models, the more sensitive variables are not only a one-year pause in a single state, but whether similar rules will spread and whether approval timelines and power supply matching will systematically raise costs.

Michigan explains that projects will move slower

New York is a landmark case, but not an isolated one. Michigan’s Saline Township project shows that local resistance doesn’t necessarily knock a project out, but it can change how it is carried forward.

OpenAI said the local Stargate campus is 1GW in size. Blackstone and Related Digital previously announced that they would provide $16 billion in financing support for the Oracle data center project in Saline Township; participants also include OpenAI and Walbridge. The project faced resident opposition; the town council at one point rejected it. The developer then pushed the project forward through litigation and a settlement.

The signal to the market from this case is very direct. As long as the strategic value is high enough and the capital pool deep enough, companies may still be able to push projects forward through legal procedures, compensation arrangements, and design adjustments. The price is a longer timeline, with political and legal costs entering the project’s return calculation.

AI data centers are no longer just a capital expenditure competition between cloud providers and developers. They also have to pass through a filter of local governance. Residents see the trade-offs among electricity, water, noise, roads, and tax revenue. Companies see orders delivered, compute capacity coming online, and customer commitments.

New Mexico exposes hard constraints on energy support

New Mexico’s Project Jupiter is closer to an energy bottleneck. In March, the state’s land commissioner denied Energy Transfer’s application to cross state trust lands with a pipeline, and in July it also denied the company’s request for reconsideration. The pipeline was originally intended to serve Oracle-related data center projects; the denial reasons involved greenhouse gas emissions and water resource pressure.

This detail is harder than resident opposition. Data centers don’t run just because servers arrive; they require stable power, transmission interconnection, and cooling systems. If any permitting step gets stuck, AI capital expenditures turn from a financial plan into a queued engineering project.

On July 1, Oracle said the project’s power方案 has been adjusted to Bloom Energy fuel cells. This indicates that large data centers will continue to look for alternative energy configurations, but alternative solutions also mean the costs, technology, and delivery time must be reassessed.

Cases like this can affect investors’ judgment of the quality of AI infrastructure companies. Getting servers and chips is just the first step; being able to connect the equipment to a stable, compliant energy system that is acceptable to local areas determines whether the project can turn from an announcement into revenue.

Power and site selection capabilities are being repriced

The market implication of this round of change is not that AI expansion ends, but that the winner criteria are changing. Chip supply is the first-layer threshold; power, site selection, and permitting ability are becoming the second-layer moat.

According to publicly reported statistics, in 2025 the United States had about $156 billion worth of data center projects affected by local opposition, lawsuits, or suspensions. In 2026 Q1, another roughly $130 billion in projects were added to the blocked or delayed range. These numbers do not mean all projects are fully canceled; a significant portion may only be postponed, relocated, or renegotiated, but they are enough to show that resource friction has already affected the pace of capital deployment.

Under pressure are cloud providers and developers advancing projects in regions with tight grids, strict regulation, and sensitive water resources. Oracle, Microsoft, Google, Amazon, and infrastructure-capital players like Blackstone all need to reassess the permitting certainty of where projects are located and their power-supply方案.

The directions that benefit are also clearer. States with excess power and policy-friendly environments—stable power sources such as natural gas and nuclear power, dedicated power interconnection, technologies like liquid cooling and closed-loop cooling that reduce water use—will gain higher strategic weight. Power is becoming a prerequisite for whether capital expenditures can be realized.

Project migration speed determines valuation realization

If New York can form an executable environmental assessment and community investment framework within one year, the pause order is more like a reset of rules. Project costs rise and construction time lengthens, but the compliance path becomes clearer. If the standards are too strict, or if other states follow with even stronger restrictions, short-term friction could potentially turn into a nationwide bottleneck.

Whether Trump pushes for federal-level support for energy and AI infrastructure will also affect this tug-of-war. Even so, grid interconnection, local water resources, and residents’ electricity prices will not be directly smoothed out by political statements.

AI data centers will still be built, but starting now, whoever can get power, approvals, and community acceptance faster is more likely to turn AI capital expenditures into real revenue.

Click to learn more about Lyting BlockBeats hiring opportunities

Welcome to join Lyting BlockBeats’ official community:

Telegram subscription group: https://t.me/theblockbeats

Telegram discussion group: https://t.me/BlockBeats_App

Twitter official account: https://twitter.com/BlockBeatsAsia