Liquid Cooling for AI Act of 2025

Bill Number

S. 3269

Origin Chamber

Senate

Congress

119th Congress, Session 1

Policy Area

Energy

Status

Introduced

Latest Action

2026-04-15: Committee on Energy and Natural Resources Subcommittee on Energy. Hearings held.

Last Updated

2026-04-16T11:03:25Z

AI-Generated Summary

Purpose

The Liquid Cooling for AI Act of 2025 aims to address the growing energy demands of artificial intelligence (AI) and high-performance computing by directing an independent review of liquid cooling technologies. Liquid cooling uses liquids to remove heat from electronic components more efficiently than traditional air cooling, which is becoming insufficient for high-density AI systems. The legislation seeks to inform federal decision-making on deploying these technologies in government facilities to improve efficiency, reduce energy use, and support U.S. leadership in AI.

Key Provisions

• Findings by Congress: Highlights rapid growth in data center electricity use (from 1.9% of U.S. total in 2018 to 4.4% in 2023, projected to reach 6.7–12.8% by 2028 due to AI and cloud computing). Emphasizes the need for liquid cooling to handle heat from AI hardware, including types like direct-to-chip cooling (liquid touches specific hot parts like processors) and immersion cooling (components submerged in fluid). Notes benefits such as better performance, higher density, and opportunities for heat-reuse (capturing waste heat for other uses, like heating buildings).

• Definitions: Provides clear terms, such as:
• AI: Refers to systems that perform tasks requiring human-like intelligence (as defined in existing law).
• Single-phase cooling: Fluid stays liquid throughout the process.
• 2-phase cooling: Fluid changes to gas to absorb heat.
• Heat-reuse: Redirecting waste heat for useful purposes.

• GAO Review (Led by the Government Accountability Office, an independent watchdog agency): Must start within 30 days of enactment and cover:
• Research needs, costs, and benefits of liquid cooling for high-performance computing.
• Energy savings, including reduced need for new power lines or infrastructure.
• Impacts on computing power, performance, reliability, and cybersecurity.
• Market trends over the past 5 years and future projections.
• Comparisons of cooling methods (e.g., direct-to-chip vs. immersion) on factors like efficiency, safety, maintenance, and long-term costs.
• Coolant options (e.g., water-based or specialized fluids), system components (e.g., pumps, filters), and failure risks (e.g., leaks).
• Heat-reuse opportunities and mitigation strategies for issues like corrosion or leaks.
• Recommendations on adopting liquid cooling over air cooling, federal research, best practices, security enhancements, and training.

• Stakeholder Involvement: GAO must consult federal, state, local governments, private sector, academia, and National Laboratories (government research facilities). An advisory committee, including experts from industry (e.g., data center operators, hardware makers), will assist and terminate after the Department of Energy (DOE) report.

• Reporting Requirements:
• GAO submits a report with findings and recommendations to DOE and key congressional committees (Senate Energy and Natural Resources; House Science, Space, and Technology; House Energy and Commerce) within 90 days of enactment.
• DOE assesses the GAO report and submits its own to Congress within 180 days, including considerations for U.S. AI leadership and research recommendations on liquid cooling and heat-reuse.

Significant Changes to Existing Law

This bill introduces a new mandate for a targeted technology assessment, which does not amend prior laws but builds on existing frameworks like the National Artificial Intelligence Initiative Act of 2020 (defining AI) and the Energy Policy Act of 2005 (defining National Laboratories). It creates a structured process for evaluating emerging cooling technologies, potentially influencing future federal energy and AI policies without directly altering current regulations.

Potential Impacts

• On Government Agencies: Requires GAO and DOE to conduct reviews, informing efficient AI deployment in federal facilities (e.g., reducing energy costs and infrastructure needs). Could lead to standardized best practices for government data centers, promoting energy efficiency and innovation.

• On Citizens: Indirect benefits through lower overall energy consumption by data centers, potentially stabilizing electricity costs and reducing environmental strain from power generation. May enhance national cybersecurity and computing reliability for public services relying on AI.

• On International Relations: Supports U.S. competitiveness in global AI development by addressing technical barriers to advanced computing, helping maintain technological leadership amid international competition (e.g., from China in AI infrastructure).

Main Stakeholders Affected

• Federal Government: GAO (leads review), DOE (assesses and advises on research), National Laboratories (provide expertise), and other agencies deploying AI systems.
• Congress: Specified committees oversee and receive reports to guide future legislation on energy and technology.
• Private Sector: Data center operators, hardware manufacturers, fluid producers, and AI developers (consulted for input and affected by recommended standards).
• Academia and Research Institutions: Involved in consultations to shape research priorities.
• State/Local Governments and Public Interest Groups: Provide input on regulatory and practical implementation, potentially influencing local energy infrastructure.

Notable Legal, Constitutional, or Political Implications

• Legal: Establishes a temporary advisory committee under federal guidelines, ensuring transparency through required consultations and reports. No enforcement mechanisms beyond reporting, but recommendations could spur voluntary standards or future binding regulations on data center efficiency.

• Constitutional: Aligns with Congress's authority to oversee federal spending and technology policy (e.g., commerce clause for interstate energy use). No apparent conflicts with free speech, privacy, or due process, as it focuses on technical assessment rather than restricting activities.

• Political: Bipartisan introduction (by Senators McCormick, Coons, Budd, and Schiff) signals broad support for AI advancement and energy innovation. Highlights national security and economic priorities in AI, potentially influencing debates on federal R&D funding and climate goals by addressing data center emissions indirectly through efficiency gains.

This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.

Sponsor

Sen. McCormick, David [R-PA]

Cosponsors (3)

Sen. Coons, Christopher A. [D-DE], Sen. Budd, Ted [R-NC], Sen. Schiff, Adam B. [D-CA]

Recent Actions

• 2026-04-15: Committee on Energy and Natural Resources Subcommittee on Energy. Hearings held.
• 2025-11-20: Read twice and referred to the Committee on Energy and Natural Resources.
• 2025-11-20: Introduced in Senate

Bill Versions

• Liquid Cooling for AI Act of 2025 — issued 2025-11-20 — PDF (11 pages)