Nationwide Pause on AI Data Centers

Summary#

This bill would pause building or upgrading large, high‑power data centers used for artificial intelligence (AI) across the United States. The pause would last until Congress passes new laws that set safety reviews for AI products, protect workers and communities, and explicitly end the pause. It also adds quarterly public reporting on data centers by the Department of Energy (DOE) and sets export limits on AI‑related computing hardware to countries without similar safeguards. The stated goal is to slow AI expansion until protections are in place for safety, jobs, energy bills, and the environment.

Key changes:

• Halts construction and upgrading of “AI data centers” nationwide starting at enactment, until new federal laws meeting specific protections are passed and the moratorium is expressly ended.
• Defines “AI data center” to include sites used to build or run AI models at scale, or very large/high‑density facilities (over 20 megawatts with high‑power racks or liquid cooling), even if multi‑building and under common control.
• Requires DOE to publish quarterly reports on each AI data center, including water and energy use, emissions, wastewater heat, cooling chemicals, noise, jobs and wages, financing, land/utility/government deals, and a certification about subsidies.
• Authorizes DOE to subpoena information, require written responses, conduct inspections, and tie future federal permitting to compliance.
• Directs the Commerce Department to bar exports and transfers of computing hardware for AI data centers or large‑scale AI training/deployment in countries that lack comparable AI safety, labor, and community/environment protections.

What it means for you#

• Businesses that build or operate data centers

  • You would have to stop starting or continuing construction or upgrades of facilities that meet the bill’s “AI data center” definition until Congress passes new laws that meet the listed conditions and explicitly lift the pause.
  • Routine “upgrades” are not defined in the bill; it is unclear whether hardware refreshes, density increases, or efficiency retrofits would be allowed.
  • You would need to provide detailed information for DOE’s quarterly public reports and may face subpoenas or inspections if you do not.
  • After the moratorium ends, any new or upgraded AI data center would need to meet added conditions, including: not raising consumer utility bills, not worsening climate change or harming the environment, winning affected community approval, receiving no government subsidies, and creating union jobs with strong labor standards.

• AI developers, cloud providers, and large tech firms

  • Expansion of U.S. AI compute capacity in high‑power facilities would be paused. This could delay new capacity for training or running large AI models.
  • Export controls would likely restrict sending AI‑oriented chips and related hardware for use in AI data centers or large‑scale AI work to many countries that do not have similar laws, limiting overseas deployments and sales for those uses.

• Hardware makers and exporters

  • You could face new export bans for semiconductors, computers, networking gear, and storage if they are intended for AI data centers or large‑scale AI uses in countries without comparable protections. Compliance and end‑use screening would likely become stricter.

• Construction trades and unions

  • New data center construction and upgrade projects that meet the definition would pause. If and when the moratorium lifts, qualifying projects would be required to use union labor standards, including prevailing wages, registered apprenticeships, and project labor agreements.

• Utilities and local communities

  • Proposed high‑power AI data center projects would pause. The bill signals that, after the pause, such projects must not raise consumer energy bills, must not harm the environment, and must get affected community approval.
  • DOE’s public reports would disclose local water use, energy demand, emissions (including fenceline air monitoring results), wastewater heat, and noise from each AI data center.

• Existing data centers

  • Ongoing operations could continue, but upgrades are paused if your site meets the bill’s definition.
  • You would be subject to DOE’s quarterly reporting. The bill requires each report to include a “certification” that the center has not used federal, state, or local subsidies; how this applies to existing centers that did receive subsidies is unclear.

• Federal agencies

  • DOE would create and publish quarterly data center reports and enforce information requests.
  • The Commerce Department (through export controls) would implement new end‑use and destination bans tied to AI data centers and large‑scale AI work in countries without comparable laws.

Expenses#

The bill may increase administrative and enforcement costs for the federal government, but no estimate is available.

• DOE would face new, ongoing costs to collect, verify, and publish detailed quarterly reports for each covered data center, including possible inspections and subpoenas.
• The Commerce Department would face costs to design, implement, and enforce new export restrictions based on end‑use and the “comparability” of foreign laws.
• Businesses would likely face compliance costs to provide data for DOE reports and to meet export screening and documentation requirements.
• Construction pauses could create private costs from delays or halted projects, but no public fiscal estimate is provided.
• No appropriations or specific funding amounts are identified in the bill text.

Proponents' View#

• The bill appears intended to prevent rapid AI expansion from outpacing public safety, civil rights, and workplace protections by pausing new high‑power AI facilities until safeguards exist.
• It could protect consumers from higher electricity bills by requiring that future AI data centers not raise utility costs.
• It aims to reduce environmental harms by requiring that future centers not worsen climate change or damage local environments and by making detailed environmental data public.
• It could give communities a say by requiring local approval for future AI data center projects.
• It seeks to ensure workers share in AI gains by requiring union jobs with strong labor standards on future projects and calling for broader policies to prevent job loss and share wealth.
• Export controls could reduce the risk that AI development simply shifts to countries without comparable protections.

Opponents' View#

• One concern is that the moratorium is open‑ended and only lifts if Congress passes broad new laws and explicitly ends the pause, which could freeze new AI data center capacity for an extended period.
• The definition may capture some non‑AI high‑density or high‑power data centers, potentially pausing projects not primarily focused on AI.
• “Upgrading” is not defined; it is unclear whether maintenance, efficiency retrofits, or hardware refreshes would be blocked.
• DOE’s reporting requirements are extensive (including financing, labor data, and detailed environmental metrics). This may be burdensome and raises questions about confidentiality and how data will be verified, especially for items like fenceline air monitoring that some facilities may not currently conduct.
• The bill requires DOE reports to include a certification that each data center has not used any subsidies, but it only bans subsidies for future projects after the moratorium. How this applies to existing centers that have received incentives is unclear.
• Export bans tied to end‑use and to whether other countries have “comparable” laws may be hard to implement, could restrict U.S. hardware sales, and may affect international competitiveness if criteria for “comparable” are unclear.