Energy and water use reporting for AI data centers

In short

No federal statute requires privately owned AI data centers to report their energy or water consumption to any federal agency. Durbin press release Several bills introduced in 2025 and 2026 aim to create such mandates. None have become law. In the meantime, a wave of state laws is filling the gap. In 2025 alone, more than 190 state bills addressing AI data centers were introduced nationwide and more than two dozen were enacted. Roughly 40 of those bills required some form of energy or environmental disclosure, and about 30 addressed water consumption. Trade press analysis On the voluntary side, the ENERGY STAR program, the GHG Protocol, and investor pressure are pushing operators toward more transparency even where no law requires it. The upshot is a patchwork of obligations that changes quickly and depends on where a facility sits and who owns it.

What does current federal law require?

The short answer, nothing binding on private operators

There is no federal statute that makes a privately owned AI data center measure, track, or report its energy use or water use to a federal agency. CRS, CRS report Federal reporting obligations apply only to federal agencies for their own facilities.

The main federal statute that mentions AI data centers is the Energy Independence and Security Act of 2007. It defines a data center as any facility that mainly houses electronic equipment for processing, storing, and transmitting digital information. The building can be free standing or part of a larger structure. It may be a facility within a larger structure that uses environmental control equipment to maintain the proper conditions for the operation of electronic equipment. 42 U.S.C. § 17112(a)(1)(B), 42 U.S.C. § 17112(a)(1) That definition is used in several of the proposed bills, but it does not impose a reporting duty on its own.

The Energy Act of 2020 directed the Department of Energy to study data center electricity use and created a Data Center Energy Practitioner training program. Federal agencies must consider having their own data centers evaluated once every four years. Pub. L. 116-260, Div. Z, § 1003 That program does not reach private facilities.

The Energy Policy and Conservation Act lets DOE set energy conservation standards for commercial air conditioning equipment and distribution transformers used in AI data centers. But the compliance duty falls on the manufacturer at the time of shipment, not on the operator as an ongoing reporting requirement. 42 U.S.C. § 6313(a), 42 U.S.C. § 6317(a), 42 U.S.C. § 6316(b)(1), 42 U.S.C. § 6302(a)(5), 42 U.S.C. § 6317(f)(1)(B), 42 U.S.C. § 6317(f)

The Federal Data Center Enhancement Act of 2023, implemented through an OMB memorandum in January 2025, requires each federal agency to report on compliance with OMB’s data center requirements and to have new or substantially upgraded data centers evaluated by specialists certified in data center energy efficiency. Again, private AI data centers that solely support commercial cloud services to a wide variety of customers are presumptively not covered, though contractor-operated AI data centers on behalf of agencies are covered. OMB M-25-03

A procedural hurdle, the Paperwork Reduction Act

Even if a federal agency wanted to collect energy or water data from private AI data centers, it would first need approval from the Office of Management and Budget under the Paperwork Reduction Act. That process requires public notice and comment and can take months. 44 U.S.C. §§ 3501 et seq. This is the same process that slowed the Energy Information Administration’s earlier efforts, discussed below.

Proposed federal bills that would impose reporting

Congress has introduced several bills that would directly require AI data centers to report electricity and water usage. As of May 2026, none have been enacted.

Clean Cloud Act of 2025 (S. 1475)

This bill, introduced by Senators Whitehouse and Fetterman, would amend the Clean Air Act to give EPA and the Energy Information Administration authority to collect data from any AI data center or cryptocurrency mining facility drawing more than 100 kilowatts of power. Each covered facility would report its total electricity consumption, the identity of its electricity provider, any power purchase agreements, and the mix of energy sources used. EPA would then calculate each facility’s greenhouse gas emission intensity.

The bill would set regional emission performance standards that start at the current grid average intensity and decline by 11 percent each year until they reach zero in 2035. A facility that exceeds the standard would pay a fee of $20 per excess ton of CO₂ equivalent, adjusted annually for inflation plus an additional $10. S. 1475, 119th Cong. A companion bill was introduced in the House. Both were referred to committee.

Data Center Water and Energy Transparency Act of 2026 (S. 4213)

Introduced by Senator Durbin in March 2026, this bill would require AI data center operators and developers to report the energy and water consumed at each facility to the state where it sits. A proposed facility would have to submit estimated consumption for its first five years. States would make the information available to local governments, aggregate and anonymize the data, and submit it to EPA, DOE, and the U.S. Department of Agriculture. Those three agencies would jointly issue regional reports. Facilities that do not comply could be fined. S. 4213, 119th Cong., Sen. Durbin press release

Data Center Transparency Act (H.R. 6984)

Introduced by Representative Menendez in January 2026, this bill would require EPA to submit quarterly public reports on total water consumed by U.S. AI data centers, water reuse amounts, effects on local water systems (including potable water availability, demand on utilities, service disruptions, residential rate changes, and pollutants discharged), total greenhouse gases emitted, and cumulative effects on overburdened communities. The EIA would collect energy consumption data every six months from every U.S. AI data center and publish semiannual reports showing total energy consumption by state, changes over time, the number of new AI data centers, and effects on household energy bills. The bill uses the statutory definition of a data center from the 2007 Energy Act. H.R. 6984

AI Data Center Site Selection Transparency Act of 2026

Introduced by Representative McIver in April 2026, this bill would require developers to publicly disclose the location of a proposed AI data center at least 180 days before taking any definitive step toward development. Developers would also have to disclose electricity use, water consumption, cooling demands, and environmental impacts backed by independent third party analysis. The bill would restrict nondisclosure agreements with public entities and treat violations as unfair or deceptive practices under the FTC Act. Rep. McIver press release

Other federal proposals

Several other bills address AI data center impacts without directly imposing reporting on operators. The Unleashing Low-Cost Rural AI Act would mandate an interagency study on how AI data center expansion affects rural energy supply, reliability, and consumer costs. The Securing Reliable Power for Advanced Technologies Act would accelerate federal permitting of energy infrastructure for certain AI facilities. The Unleashing Low-Cost Rural AI Act does not create a reporting obligation for AI data center operators, but the Clean Cloud Act (S. 1475) would require them to report electricity consumption data to EPA and the Energy Information Administration. Route Fifty, CRS R48762

State reporting requirements are already taking shape

While Congress debates, state legislatures are moving. In 2025, more than 190 bills addressing AI data centers were introduced across all 50 states. More than two dozen were enacted. About 40 of those enacted or proposed bills required some form of energy use or environmental impact disclosure. Approximately 30 bills addressed water consumption directly. Trade press analysis The table below is a sample of enacted and pending measures, not an exhaustive list.

[Table, state reporting requirements sample]

Bills that did not become law are also instructive. California Governor Newsom vetoed AB 93, which would have required AI data centers to report water use estimates before obtaining a business license. The veto message cited concern about imposing rigid reporting requirements on an industry of global and regional importance. Gov. Newsom veto message New Jersey Governor Murphy vetoed S4293, which would have required energy and water usage reports to the Board of Public Utilities, recommending that data be anonymized to protect proprietary information. Gov. Murphy veto message Virginia Governor Youngkin vetoed HB 1601, which would have required site assessments for high-energy-use facilities such as AI data centers, calling it a one size fits all approach that would create unnecessary red tape. HB 1601 veto

A handful of pending bills go much further. South Carolina HB 4583 would require closed loop cooling systems with zero net water withdrawal and no municipal water use. Kansas SB 400 would also require closed loop cooling and allow municipalities, local district attorneys, and the state attorney general to sue for injunctions against violators. MultiState

The state landscape is moving fast. AI data center operators and developers must track each jurisdiction individually because the obligations vary widely.

Voluntary programs and private standards

ENERGY STAR certification

The joint EPA DOE ENERGY STAR program offers a voluntary certification for AI data centers. To earn it, a facility must score 75 or higher on a 1 to 100 performance scale, placing it in the top 25 percent of similar facilities. The score is based on Power Usage Effectiveness, or PUE, defined as total facility energy divided by IT equipment energy. Over half the building’s floor area must be used for AI data center operations. Energy data for IT loads and the whole building must be gathered for 12 consecutive months, professionally verified, and submitted through EPA’s Portfolio Manager tool. Certification lasts one calendar year and must be renewed annually. There is no fee. ENERGY STAR, Data Center Knowledge

The program does not require water use reporting. As of September 2024, only 270 of the roughly 5,381 U.S. AI data centers held the certification. HWG LLP

GHG Protocol updates

The Greenhouse Gas Protocol, governed by the World Resources Institute and the World Business Council for Sustainable Development, is the main voluntary standard for reporting emissions. It is used by 97 percent of S&P 500 companies that report to CDP and is referenced in California SB 253 and the EU’s Corporate Sustainability Reporting Directive. Trellis, GHG Protocol on SB 253, GHG Protocol on CSRD and SB 253

In October 2025, the GHG Protocol opened a public consultation on proposed updates to its Scope 2 Guidance, the standard for reporting emissions from purchased electricity. The key proposed changes include hourly matching of renewable energy credits to actual consumption, a deliverability requirement that contractual instruments must come from generators that could plausibly deliver electricity to the consumer on a connected grid, a new hierarchy of emission factors that values spatial and temporal granularity, and a legacy clause for existing contracts. The comment period runs through January 31, 2026. Implementation would be phased over several years. GHG Protocol, PwC

The Scope 3 Standard is also being revised, with a draft expected for consultation in 2026 and a final standard in 2027. The Corporate Standard follows a similar timeline. Corporate Standard Development Plan, Scope 3 Standard Development Plan

These revisions matter for AI data center operators because they will change how renewable energy purchases are accounted for and how supply chain emissions from hardware manufacturing and construction are calculated. GSI Environmental, GHG Protocol

Voluntary corporate reporting and the patchwork of data

The largest cloud providers already report some environmental data, but the level of detail varies widely.

• Meta reported a fleet wide PUE of 1.08 in 2024 and reported that its total water use rose 51 percent from 2020 to 2024, reaching 5,637 megaliters. Meta reports water use for both owned and leased sites. Water used during construction is reported separately and excluded from total water withdrawal. Meta 2025 Environmental Data Index
• AWS reported a global PUE of 1.15 in 2024 and a water use effectiveness metric of 0.15 liters per kilowatt hour. It does not report total water consumption, only the per unit figure. AWS 2024 Sustainability Report
• Google reports water use for owned and leased sites but not for third party operated ones. One facility in Council Bluffs, Iowa alone consumed about 1.0 billion gallons of potable water in 2024, roughly 2.8 million gallons per day. Google 2025 Environmental Report
• Microsoft reported total water consumption but not by site. Its total emissions across Scope 1, 2, and 3 rose 23.4 percent from its 2020 baseline, even as energy use climbed 168 percent. Microsoft 2025 Sustainability Report

These voluntary disclosures are not standardized. A developer or a utility commission cannot compare water use across operators on an apples to apples basis from the numbers released today.

Investor pressure adds another layer

Institutional investors are pushing for more granular, site level data. In the run up to 2026 annual meetings, more than a dozen investors, including Trillium Asset Management and Calvert Research and Management, filed shareholder resolutions or engaged with Amazon, Microsoft, Alphabet, and Meta seeking site specific water and energy consumption data. Trillium’s 2025 resolution at Alphabet won support from nearly 25 percent of independent shareholders. Reuters

For a developer or operator, this means that even in a state with no reporting law, a large customer, a lender, or an equity partner may demand detailed energy and water data as a condition of financing or a power purchase agreement.

The gaps and the obstacles

The EIA’s failed pilot

In 2021, the Energy Information Administration surveyed 50 AI data center buildings for a pilot study. It completed 12 interviews for a response rate of 26 percent, but only 9 were confirmed AI data centers. Only 1 of 20 private AI data centers responded. The item nonresponse rate was 90 percent for the number of servers and 80 percent for PUE. The agency concluded that AI data center estimates are unlikely to be feasible with its current methods. EIA 2018 CBECS Data Center Pilot Results

In 2026, EIA launched three new voluntary pilot surveys in Texas, Washington state, and the Northern Virginia and Washington, DC region, targeting 196 companies operating 1,066 facilities. The surveys cover energy sources, consumption, site characteristics, server metrics, and cooling systems. After the pilots, scheduled for completion by late September 2026, EIA plans to develop a mandatory nationwide survey. IER, WIRED The fate of that mandatory survey depends on whether Congress gives EIA statutory authority and whether the Paperwork Reduction Act process can produce a workable collection instrument.

Nondisclosure agreements hide water use

In several states, local utilities are barred from releasing data about AI data center water consumption because of nondisclosure agreements with operators. The American Water Works Association reported this problem. In Georgia, a public records request by a television station resulted in water usage information being redacted. That episode prompted the introduction of Georgia SB 421 to prohibit such NDAs. MultiState

Clean Water Act does not regulate quantity

The federal Clean Water Act regulates water quality, not water quantity. It does not give any federal agency authority to limit how much water an AI data center can withdraw. E&E News That gap leaves water allocation entirely to state law, which varies widely. Louisiana, for example, requires registration for groundwater pumping only if the user plans to tap more than 50,000 gallons per day, with no permit required. E&E News

Industry opposition

The Data Center Coalition, the principal trade association, has opposed mandatory reporting requirements. The coalition’s vice president of state policy has argued that they unfairly penalize one industry. Tech giants argue that water use data could be used by competitors to infer facility design. E&E News

Practical takeaways for counsel and developers

The current legal environment is a patchwork with no single source of truth. A developer or operator should consider the following steps.

1. Map each project against the state laws where it sits. A facility in Minnesota faces different rules than one in Texas, even if the same operator owns both. The table above is a starting point, not a substitute for a state specific review.
2. Plan for disclosure in project documents even where no law requires it. Lenders, investors, and large customers are increasingly asking for site level water and energy data. A project that cannot produce that data may be harder to finance.
3. Treat the voluntary GHG Protocol and ENERGY STAR standards as a baseline. They are becoming embedded in mandatory rules, like California SB 253, and in investor demands. Aligning data collection now saves cost later.
4. Watch the EIA pilot and the proposed federal bills. If a mandatory federal survey or an enacted Clean Cloud Act emerges, the reporting burden will shift from a patchwork of state obligations to one federal standard with penalties.
5. Review nondisclosure agreements with utilities. Laws like Georgia SB 421 may void contractual secrecy around water use. An operator should assess whether existing NDAs could become unenforceable.
6. Consider closed loop cooling systems early. Pending bills in South Carolina and Kansas would require them. While those bills are not law, they signal a direction that could affect site selection and permitting in water stressed regions.
7. Keep the large numbers in mind for siting decisions. North American AI data centers consumed nearly 1 trillion liters of water in 2025. Reuters A 100 MW facility uses water equivalent to about 2,600 households, and about 80 percent of the water drawn for cooling evaporates. CRS report, NASUCA Communities are starting to ask hard questions, and permitting will be harder where water is scarce.

Key takeaways

• No federal mandate exists today. Privately owned AI data centers do not report energy or water use to any federal agency as a matter of federal law.
• Several federal bills would change that. The Clean Cloud Act, the Data Center Water and Energy Transparency Act, and the Data Center Transparency Act would each impose reporting and, in the case of the Clean Cloud Act, emission standards and penalties. None have passed.
• State laws are the real action. More than two dozen bills were enacted in 2025 alone, with about 40 bills (enacted or proposed) addressing energy disclosure and approximately 30 addressing water. Minnesota, Texas, Iowa, and Virginia all have reporting obligations on the books.
• California SB 253 is the broadest in scope. It requires large companies to report Scope 1 and 2 emissions starting in 2026, reaching AI data center operators with over $1 billion in revenue.
• Voluntary programs set market expectations. ENERGY STAR certification and the GHG Protocol are used by investors and large customers as de facto standards, even where not required by law.
• NDAs around water use are being challenged. Georgia, New Jersey, and other states are moving to prohibit secret water deals, and existing NDAs may become unenforceable.
• Investor pressure is mounting. Shareholder resolutions seeking site level data are gaining support, and major asset managers are pushing hyperscalers for more granular disclosure.
• EIA and DOE are building the infrastructure for mandatory federal data collection. The 2026 pilot surveys are a first step. A mandatory nationwide survey is planned. Whether it becomes law depends on Congress.
• A facility’s water consumption is becoming a material site selection and permitting factor. In water stressed regions, closed loop cooling mandates and aquifer testing requirements are on the rise, and community pushback is growing.

Frequently asked questions

Q:Is there a federal law requiring AI data centers to report their electricity use?

A:
No. As of May 2026, no federal statute imposes an energy reporting obligation on privately owned AI data centers. The Federal Data Center Enhancement Act applies only to federal agency facilities. CRS R48646

Q:What would the Clean Cloud Act require?

A:
The Clean Cloud Act of 2025 would require any AI data center or cryptomining facility having more than 100 kW of installed information technology nameplate power to report electricity consumption, power purchase agreements, and fuel mix to EPA. EPA would calculate each facility’s emission intensity and set declining performance standards, with fees for excess emissions. The bill has not been enacted. S. 1475, 119th Cong.

Q:Which states already require AI data center energy or water reporting?

A:
Several states have enacted laws. Minnesota requires water consumption disclosure and aquifer testing for large facilities. Iowa requires annual reporting of backup fuel and electricity purchases. Texas requires interconnection applicants to disclose other applications and backup generation. Virginia requires water utilities to report volumes delivered to AI data centers and requires water consumption estimates for proposed projects. California SB 253 will require Scope 1 and 2 emissions reporting starting in 2026 for large companies. This list is not exhaustive.

Q:Does the ENERGY STAR program require water reporting?

A:
No. ENERGY STAR certification for AI data centers is based on energy efficiency, measured by PUE. It does not require any water use data. ENERGY STAR Score for Data Centers, ENERGY STAR Technical Reference

Q:What is the GHG Protocol and why does it matter?

A:
The GHG Protocol is the most widely used voluntary standard for measuring and reporting greenhouse gas emissions. It is required by California SB 253 and referenced by the EU’s CSRD. Proposed updates to its Scope 2 Guidance, including hourly matching of renewables, could change how AI data centers account for clean energy purchases. GHG Protocol

Q:Why did the EIA’s earlier data collection effort fail?

A:
The EIA’s 2021 pilot survey of 50 AI data centers received only 9 usable responses. Only 1 private operator responded. Security concerns, difficulty finding knowledgeable contacts, and the voluntary design made the data unusable. The agency is now running new pilot surveys in three regions. EIA press release, CRS report, EIA press release

Q:Can a state prevent a water utility from disclosing how much water an AI data center uses?

A:
Georgia is moving to prohibit nondisclosure agreements that hide water usage from the public. SB 421 would bar local governments, authorities, and political subdivisions from entering into nondisclosure agreements regarding water or electricity usage of any entity. Some states are considering legislation that would restrict nondisclosure agreements between AI data center operators and local governments regarding water and energy usage. MultiState

Q:What should a developer do to prepare for new reporting laws?

A:
A developer should map the specific state laws that apply to each project, build energy and water data collection into the design from the start, and align with the GHG Protocol and ENERGY STAR where possible. Anticipating investor and lender data requests now can reduce the cost of scrambling later.

Q:What is the SEC’s climate disclosure rule status?

A:
The SEC adopted a rule in March 2024 that would have required large companies to report Scope 1 and 2 GHG emissions. The rule was stayed, and the SEC voted in March 2025 to end its defense. The rule is not in effect and is unlikely to be enforced. SEC Press Release 2024-31, SEC Press Release 2025-58

Subscribe to The Compute Law Brief

The Compute Law Brief is a free weekday newsletter on the law of AI infrastructure across tax, real estate, construction, power, and deals. The big US build markets and federal law. Three minutes a morning. No paywall, and no email gate to read the blog. Subscribe if you want it in your inbox.

Junde Liu, JD, LL.M. (Taxation) candidate at UF Law. Originally published on Compute Law Blog. This article is general information and does not constitute legal advice. Reading it does not create an attorney client relationship. The reader should not act on the basis of any content here without first consulting a licensed attorney in the relevant state. Last reviewed for accuracy May 23, 2026.