AI Industrial Strategy Under Trump 2.0
How environmental rollbacks, grid expansion, and public subsidies are reshaping U.S. policy for data center growth
![Constellation Energy [CC]](https://commons.wikimedia.org/wiki/File:Exelon_Three_Mile_Island_Nuclear_Generating_Station_(cropped).jpg){kind=link}
The Trump administration’s AI industrial strategy repurposes U.S. energy, environmental, and land-use policy to fast-track infrastructure buildout, using national security rhetoric to justify regulatory rollbacks while offloading the social, financial, and ecological costs onto the public.
Last week, I wrote about the U.K.’s AI Growth Zone policy and how it reshapes infrastructure development through a process I described as a new form of enclosure. Since then, I’ve been examining how similar dynamics are emerging in the United States. This week, I focus on recent shifts in U.S. energy and environmental policy, where the drive to accelerate AI infrastructure is prompting significant changes to permitting processes, environmental oversight, and infrastructure approvals. These policies, framed as efforts to streamline regulation and promote national competitiveness, increasingly enable private actors to capture the gains of expansion while leaving the social and ecological costs to be absorbed by the public.
The political logic behind this strategy was made especially clear last Friday, when Secretary of the Interior Doug Burgum argued that the United States must expand fossil fuel use to “win the AI arms race against China,” calling this contest the nation’s true “existential threat.” In doing so, Burgum openly derided climate action, dismissing efforts to phase out coal and invest in renewable energy as dangerous distractions that weaken the grid and jeopardize U.S. competitiveness—suggesting instead that emissions reduction policies represent a greater risk than global warming itself. This framing crystallizes how the pursuit of AI dominance is now being wielded to justify a profound rollback of climate commitments and environmental protections in the name of industrial expansion.
Burgum’s remarks came a day after the White House Office of Science and Technology Policy (OSTP) published more than 10,000 public comments on its draft national AI strategy, the so-called AI Action Plan. Spanning nearly 18,500 pages, these comments—submitted by tech firms, environmental advocates, civil rights groups, venture capitalists, and local governments—offer a striking snapshot of the policy fault lines now defining the U.S. AI landscape. While many of the submissions focus on familiar questions about copyright, algorithmic bias, labor, and trade policy, buried within them (or even omitted) is a quieter but more structurally significant debate—how the energy, land, and water demands of AI infrastructure are being mobilized to justify sweeping regulatory rollbacks and state-backed industrial support.
The Data Center Coalition (DCC), a trade group representing major cloud and infrastructure companies, used its comments to argue that tariffs on server hardware and infrastructure components could slow the deployment of data centers. It also pressed for reforms to the National Environmental Policy Act (NEPA) and the Clean Water Act (CWA) to expedite permitting for energy projects and data center siting. At the same time, several environmental groups, municipal governments, and water utilities warned that the unchecked proliferation of data center facilities could severely strain regional energy grids and water supplies, exacerbating inequality and environmental degradation.
These debates arrive at a moment of sharp political realignment. Since returning to office in January 2025, President Donald Trump has aggressively dismantled the Biden-era regulatory framework for AI governance, repealing the previous administration’s Executive Order on AI, which emphasized fairness, safety, and anti-discrimination measures. In its place, Trump has advanced a nationalist-industrial vision that redefines AI development as a matter of “human flourishing,” “economic competitiveness,” and “national security.” These priorities are being operationalized not through direct subsidies alone, but through the systematic revision of energy, environmental, and land-use policies—designed to fast-track infrastructure build-out while shielding corporate actors from the costs of expansion.
In this in-depth primer, I examine how the Trump administration’s AI industrial strategy is being constructed through energy and permitting reform, as well as environmental rollback, repurposing public regulatory systems to serve private infrastructural demands. At the heart of this process is a familiar logic of privatizing the benefits of infrastructure development while socializing the costs—onto ratepayers, local communities, and the environment. Far from an abstract competition over innovation policy, the politics of energy and environment over AI expansion represents a concrete shift in how the U.S. state governs infrastructure and resource use in the digital economy under the new administration.
AI’s Energy Demands and the Infrastructure Grab
The political economy of AI is not confined to data models or algorithmic governance—it is embedded in the material systems that power and cool the infrastructure on which these technologies depend. Nowhere is this clearer than in the dramatic rise of data center energy consumption across the United States. A December 2024 Department of Energy report projected that data center electricity use could triple by 2030, accounting for as much as 10 percent of total U.S. power demand. AI workloads, particularly large-scale machine learning and training models, are the primary drivers of this surge, requiring continuous, high-capacity, and dispatchable energy.
To secure reliable power, major tech firms have embraced an aggressive siting strategy which include co-locating data centers directly at generation facilities—particularly nuclear plants and gas-fired stations—in order to bypass the transmission grid and lock in stable energy supplies at favorable rates. One of the most high-profile examples was the late 2024 proposal by Amazon to co-locate an AI data center at Talen Energy’s Susquehanna Nuclear Plant in Pennsylvania.
But this arrangement quickly drew regulatory scrutiny. In November 2024, the Federal Energy Regulatory Commission (FERC), under then-Chair Willie Phillips, rejected the Amazon-Talen deal. Republican commissioner Mark Christie, who led the opposition, argued that such arrangements allow large industrial users to bypass the grid fees that fund shared transmission infrastructure—effectively shifting the costs of grid reliability onto residential customers and small businesses. Although the vote was not a straightforward partisan split, then-Chair Willie Phillips dissented, warning that the decision could undermine U.S. competitiveness in artificial intelligence and jeopardize both national security and electric reliability.
The stakes around this dispute only grew as the political balance at FERC shifted. Following Phillips’ departure in early 2025—encouraged by the Trump administration—the commission was left evenly divided, with a pending Republican nomination poised to secure a majority. This leadership change coincided with the Trump administration’s broader effort to reshape the regulatory landscape to prioritize rapid energy infrastructure buildout for AI-driven industrial growth.
The market reaction to FERC’s rejection of the Amazon-Talen deal revealed just how tightly financial speculation had become linked to this policy trajectory. Investors had bid up the shares of power plant owners like Vistra, Talen, and Constellation Energy on the expectation that direct power sales to AI data centers would yield steady, above-market returns. But after the deal was blocked, these stocks fell sharply, exposing the fragility of the bullish narrative tying AI infrastructure growth directly to energy sector profits. More recently, nuclear energy stocks have shown particular sensitivity to news about AI-related demand, underscoring how deeply investor expectations are now entangled with the AI boom.
Yet even as FERC’s ruling disrupted the momentum behind co-location deals, it did not settle the debate. Instead, the dispute has escalated into a central battleground in the politics of U.S. industrial policy. Industry advocates continue to argue that accelerating energy infrastructure development is essential for maintaining AI leadership, framing permitting reform and grid expansion as matters of national economic security. While a broader FERC review of co-location arrangements remains underway—and Talen pursues an appeal in the Fifth Circuit—the real action has shifted toward higher-stakes federal policy debates over permitting reform, resource adequacy, and the future of AI-related energy governance.
Emergency Declarations and Deregulation by Design
The Trump administration’s response to this unfolding conflict was immediate. On January 20, 2025, Trump signed an executive order declaring a national energy emergency, citing “precariously inadequate” energy supply as an imminent threat to national security and economic competitiveness. The order empowers federal agencies to override local and state permitting processes, accelerate fossil fuel extraction on federal lands, and fast-track the build-out of new generation and transmission capacity.
Although the order avoids directly mentioning data centers, its logic aligns seamlessly with industry demands. Under the authority of the energy emergency declaration, agencies are directed to identify and use all lawful means—including the Defense Production Act and federal eminent domain powers—to expedite infrastructure projects deemed critical to national interests. In effect, this repositions AI infrastructure as a matter of strategic resource allocation, subordinating environmental review processes and state sovereignty to the acceleration of digital capitalism.
At the same time, these moves have been reinforced through congressional hearings and legislative proposals. In March 2025, the House Energy and Commerce Subcommittee on Energy held hearings that explicitly tied AI-related energy demand to the urgency of permitting reform. Subcommittee Chair Bob Latta (R-OH) and full committee Chair Brett Guthrie (R-KY) both framed permitting delays for fossil fuel and nuclear generation as existential threats to U.S. leadership—essential to “beating China” in AI development. Witnesses from PJM Interconnection, the Southwest Power Pool, and the California Independent System Operator testified to growing backlogs in generation interconnection queues, offering further justification for legislative action to weaken permitting requirements.
This policy direction builds on the failure of bipartisan permitting reform proposals to advance in the final months of the 118th Congress. Now, with unified Republican control of Congress and the executive branch, the Trump administration is pursuing an even more aggressive deregulation agenda under the pretense of securing national competitiveness.
Ratepayers, Resource Allocation, and the Socialization of Costs
While the Trump administration frames its push for accelerated infrastructure approvals (i.e. “cutting red tape”) as a strategy to “unleash innovation” and enhance national competitiveness, the material consequence is the familiar transfer of financial and ecological burdens onto the public. Nowhere is this more evident than in the politics of grid financing. The infrastructure required to meet AI-driven power demand—new pipelines, gas plants, nuclear reactors, and transmission lines—will require billions of dollars in investment. Yet rather than fully requiring data center developers to bear these costs, the current policy trajectory risks passing them onto ratepayers, deepening structural inequalities in the energy system.
This risk is baked into the financial architecture of U.S. energy regulation. Investor-owned utilities (IOUs), which serve the vast majority of American households, operate under a model that rewards capital investment through guaranteed rates of return on their asset base. A recent analysis by the American Economic Liberties Project (AELP) finds that IOUs routinely secure rates of return far above their actual cost of capital, extracting an estimated $50 billion annually from ratepayers through excess profits. This system creates a direct incentive for utilities to favor capital-intensive infrastructure expansion—such as the build-out of new generation and transmission facilities—regardless of whether these projects are the most efficient or equitable means of meeting demand. Indeed, a 2025 Berkeley Lab report found that price increases were related to increases in capital expenditures, which grew by 50% over the same period, outpacing inflation.
The emerging AI infrastructure boom plugs directly into this incentive structure. Data centers supporting AI workloads are among the most energy-intensive industrial users, driving significant new electricity demand where they are sited. Yet rather than being treated as isolated, cost-bearing projects, these facilities risk triggering system-wide investments in grid capacity—investments whose costs may be spread across the general customer base.
This dynamic is already visible in regions experiencing concentrated data center growth. In Ohio, for example, American Electric Power has warned that electricity demand in its service territory could double by 2030, largely due to new data center developments. While AEP has sought higher interconnection fees from these large-scale users, the company has also warned that, without stronger cost-allocation mechanisms, the capital expenses of new generation and transmission infrastructure could still be recovered through general customer rates—burdens borne by households and small businesses that are not driving the new demand.
The risks of this approach have already surfaced in the Mid-Atlantic region. In 2025, PJM Interconnection, the nation’s largest grid operator, postponed its capacity auction after a complaint by environmental and consumer advocacy groups—including the Sierra Club and Public Citizen—highlighted that the auction's design could lead to capacity payments significantly exceeding market prices. These elevated payments, intended to incentivize new generation capacity, risk being passed directly onto end-users if not addressed through regulatory intervention.
Recognizing these pressures, FERC launched a formal review of PJM’s tariff structure on co-location arrangements in February 2025, with a technical conference scheduled for June 4–5, 2025, to address questions of resource adequacy, market design, and cost allocation in the context of rising industrial demand. Yet the Trump administration’s broader push to accelerate project approvals and limit environmental and regulatory oversight suggests that meaningful protection against this cost-shifting remains politically tenuous.
The combination of co-location deals, fast-tracked permitting, and weakened oversight amounts to a form of infrastructural enclosure—not through privatization, but through financial engineering that channels costs onto the public. Corporate buyers driving new demand secure favorable energy deals, while the broader costs of grid expansion and reliability are offloaded onto ratepayers and communities with little say in the process. The result is a quiet redistribution of wealth—AI’s infrastructural growth is subsidized not by the firms profiting from it, but by the public.
The EPA’s Pivot to AI Industrial Policy
What makes the current moment distinct from prior deregulation cycles is the explicit mobilization of environmental agencies—most notably the Environmental Protection Agency (EPA)—in service of AI infrastructure expansion. Under the leadership of Administrator Lee Zeldin, the EPA has repositioned itself not as a regulator but as a partner to the tech industry.
In a March 2025 statement to InsideEPA, EPA spokesperson Molly Vaseliou declared that the agency’s work to streamline permitting would “go hand in hand with AI advancement and data center proliferation.” Zeldin named making the U.S. “the artificial intelligence capital of the world” as one of his core five priorities for the agency.
To achieve this, the EPA has initiated reviews of multiple Biden-era rules, including the Power Plant Carbon Dioxide Rule, the Mercury and Air Toxics Standards, and the Good Neighbor Plan, with the goal of relaxing compliance for energy generation facilities serving AI data centers. The DCC and other industry groups have also lobbied for rolling back CAA stationary engine rules that govern diesel generator use, seeking to preserve the current allowance of 50 hours of non-emergency operation for backup generators critical to data center continuity.
The deregulatory push extends beyond air and energy rules. In its comments on the AI Action Plan, the DCC has called for the development of a CWA Section 404 Nationwide Permit for Data Center Uses, arguing that current wetlands permitting processes are too slow and unpredictable. The group has also sought to limit the jurisdiction of the Army Corps of Engineers under CWA, proposing exemptions for prior-converted cropland and upland areas that might otherwise fall under federal oversight.
Beyond permitting, the EPA is aggressively advancing its National Water Reuse Action Plan (WRAP), positioning wastewater recycling—particularly the reuse of oil and gas “produced water”—as a core strategy for meeting the water demands of data centers. The agency’s projections suggest that data center cooling water needs will rival those of a mid-sized U.S. city by 2028, a scale that raises serious questions about competition for water resources in drought-prone regions like the Southwest.
The Association of Metropolitan Water Agencies (AMWA), which represents utilities serving over 160 million people, submitted comments warning that AI infrastructure growth could place severe stress on regional water systems. They called for a comprehensive assessment of water impacts and the inclusion of water efficiency and conservation measures in the AI Action Plan. Yet the Trump administration has shown little interest in such constraints, continuing to frame water reuse primarily as an enabler of industrial expansion.
Land, Water, and the New Frontier of AI Industrial Policy
Beyond energy and grid infrastructure, the Trump administration’s AI industrial strategy is also reshaping debates over land and water use. Federal policy is now explicitly steering data center development toward brownfield sites and federal lands, including areas previously subject to environmental remediation or conservation rules.
At a March 2025 House Environment Subcommittee hearing, Republican lawmakers expressed interest in leveraging the EPA’s Brownfields Program to fast-track energy and AI-related development. James Connaughton, former Chair of the Council on Environmental Quality, testified in favor of an "Approve, Build, and Comply" permitting model, advocating for pre-cleared brownfield sites to facilitate rapid infrastructure deployment while maintaining post-construction compliance monitoring. Industry representatives discussed the potential of repurposing brownfield sites for small modular nuclear reactors and hydrogen generation facilities to provide dedicated power for data centers, aiming to minimize community opposition and streamline environmental review processes.
Water governance has emerged as an equally contentious site of policy transformation. The EPA’s National Water Reuse Action Plan, initially developed during Trump’s first term, has been revived with explicit focus on meeting the needs of AI infrastructure. Under the plan, produced water from oil and gas operations—previously considered waste—is being reframed as an industrial resource for data center cooling and other non-potable uses.
But this framing overlooks the broader implications of expanding industrial water claims in already overburdened watersheds. AMWA’s comments on the AI Action Plan include a comprehensive assessment of AI's water impacts and the integration of water efficiency and conservation measures. Without enforceable standards and regional water use planning, AI-related infrastructure could intensify competition for scarce water supplies, potentially displacing agricultural users and threatening municipal water security.
The Trump administration’s dismissal of these concerns underscores the extractive orientation of its AI industrial policy. Rather than treating water as a shared resource requiring stewardship, federal agencies are advancing policies that position water as an input for private capital accumulation.
Enclosure by Other Means
The AI Action Plan’s public comment period has surfaced a wide spectrum of views on the future of U.S. AI leadership—including serious concerns about its environmental and social costs. Yet despite the formal openness of the process, the direction of federal policy is becoming clear: under the banner of geopolitical competition and economic security, the Trump administration is positioning AI infrastructure as a strategic industrial priority, leveraging permitting reform and executive authority to accelerate corporate-led expansion.
This is not simply deregulation for its own sake. It is industrial policy through enclosure. By weakening environmental protections, eroding permitting processes, and enabling grid cost-shifting, the state is facilitating the capture of public resources—power grids, water systems, land—by private actors.
What remains conspicuously absent from this strategy is any serious mechanism for ensuring that the public, whose resources are being mobilized, will share equitably in the benefits of AI development. Instead, the costs of infrastructure expansion—higher utility bills, water scarcity, ecological degradation—are being socialized, while the profits remain concentrated.
The Trump administration’s AI industrial strategy thus reflects a broader pattern in U.S. economic governance: public goods are repurposed as growth platforms for private capital, with the language of innovation and national security deployed to justify the erosion of public oversight.
As the June FERC conference and the next stages of the AI Action Plan unfold, the question is not whether the U.S. will lead in AI, but at what price—and to whom the bill will be delivered.