The Artificial Intelligence (AI) Data Center Moratorium Act, introduced on March 25, 2026, by Sen. Bernie Sanders and backed in the House by Rep. Alexandria Ocasio-Cortez, is a proposal to pause new AI data center construction nationwide until Congress passes what the sponsors describe as strong federal AI safeguards. Those safeguards, according to contemporaneous reporting and sponsor materials, are intended to address worker and consumer protections, environmental harm, civil rights, electricity costs, and broader public-interest concerns associated with rapid AI deployment.
The bill matters because it targets the physical backbone of modern AI rather than only the software layer. Advanced models require large clusters of servers, power equipment, cooling systems, and networking infrastructure. By targeting new data centers, the proposal seeks to slow AI expansion at the point where compute capacity is actually created. That is a sharper intervention than ordinary disclosure rules or narrow sector-specific AI regulations.
Supporters frame the act as a brake on an infrastructure race that they believe is outrunning public policy. Sanders has argued that Congress has not kept pace with the scale and speed of AI change, and his office had already been calling for a federal moratorium on AI data centers before the March bill rollout. On March 25, Axios reported that the formal legislation would keep the ban in place until federal law established the safeguards the sponsors consider necessary.
The broader political backdrop is important. The Trump administration has taken the opposite general position, emphasizing faster buildout, domestic leadership, and private-sector commitments to cover power costs associated with new facilities. In early March, the White House announced a Ratepayer Protection Pledge under which major AI and hyperscale firms said they would build, bring, or buy the power needed for their data centers and pay for related infrastructure upgrades, with the administration presenting that as a way to keep household electricity prices from rising because of AI demand.
That contrast explains why the Sanders-Ocasio-Cortez bill is more than a symbolic protest. It is a direct challenge to the dominant Washington assumption that the United States must expand AI infrastructure as quickly as possible to remain economically and strategically competitive. The practical effect, if enacted, would be immediate: new AI-focused campus development, permitting, and related supply-chain commitments could stall until Congress settled the terms of federal AI law. Since Congress remains far from consensus on comprehensive AI legislation, the pause could last years rather than months.
That would almost certainly affect AI growth in the country. First, it would constrain domestic computer expansion. The International Energy Agency estimated that data centers consumed about 415 terawatt-hours of electricity globally in 2024, roughly 1.5% of global electricity use, and projected that data center electricity consumption could rise to around 945 terawatt-hours in its base case by 2030. AI is a major driver of that increase through accelerated servers and more power-dense facilities. Limiting new U.S. construction would not stop AI research outright, but it would slow the speed at which new large-scale training and inference capacity could come online domestically.
Second, a moratorium could reshape investment geography. Companies facing a halt on new U.S. facilities could redirect capital toward existing overseas campuses, non-U.S. colocation markets, or efficiency upgrades in legacy sites. That would not eliminate AI development, but it could weaken the United States’ position as the preferred location for the next wave of hyperscale and AI-optimized infrastructure. This concern is amplified by the administration’s own argument that data centers are foundational to national security and economic leadership.
Third, the bill would likely deepen ongoing disputes over energy, permitting, and rate design. The IEA has noted that hyperscale AI centers can exceed 100 megawatts of demand, while Reuters and other outlets have reported that utilities, grid operators, and major technology firms are already wrestling with connection delays, fuel choices, storage needs, and reliability pressures. In that environment, the moratorium bill effectively says the country should settle the governance question before allowing further acceleration of infrastructure. Its critics, by contrast, argue that the right answer is to build power and data infrastructure faster, not slower.
Public sentiment gives the sponsors at least some political ground to stand on. Pew Research Center reported this month that Americans are more likely to view data centers negatively than positively regarding environmental effects, home energy costs, and the quality of life for nearby residents, even though many also see economic benefits such as jobs and tax revenue. Separate Pew findings on AI more broadly show continued public wariness about the technology’s role in daily life. That does not guarantee support for a national moratorium, but it does help explain why infrastructure politics around AI are no longer confined to activist circles or local zoning fights.
Still, the measure faces steep odds. The Washington Post reported that the legislation is widely expected to fail in the current Congress, and the White House has been publicly encouraging infrastructure expansion while pressing companies to internalize more of the energy cost. Even Microsoft President Brad Smith recently acknowledged that building data centers now depends heavily on earning community trust, a sign that the industry itself recognizes the political strain around these projects.
There is also an unresolved question about scope. Public reporting indicates the bill is aimed at new AI data center construction, not at shutting down the current installed base. That distinction matters. Existing facilities would continue to power cloud services, enterprise AI deployments, and consumer applications. The likely near-term effect would therefore be deceleration, not a freeze of all AI activity. The most affected areas would be hyperscaler expansion plans, very large model training clusters, and future capacity for energy-intensive inference at the national scale.
Another reported feature of the proposal is especially consequential: multiple reports say it would also restrict exports of AI computing hardware to countries that lack comparable protections. If that element survives in formal legislative text, the bill would move beyond domestic siting policy into industrial strategy and technology trade. That would broaden the debate from local energy and labor concerns to the U.S. export-control posture and global AI competition. At the time of writing, however, broader public descriptions of the proposal remain more accessible than a complete official bill-text analysis, so that element should be treated as reported detail rather than a fully parsed legislative provision.
The Artificial Intelligence Data Center Moratorium Act is best understood as an attempt to force a national choice: should the United States keep accelerating AI infrastructure first and write the rules later, or should it pause the buildout until those rules are in place? If enacted, the bill would likely slow the development of new domestic AI capacity, delay hyperscale expansion, complicate capital planning, and strengthen arguments for a more regulated AI economy. If it fails, it will still matter because it signals that opposition to unchecked growth in AI infrastructure has moved from local resistance into mainstream federal politics. Either way, the act has already changed the conversation. The debate is no longer only about what AI can do. It is now also about how much physical infrastructure the country is willing to build before deciding which protections must come first.
