In a rare display of bipartisan urgency, United States senators from both parties have launched a coordinated legislative assault on the unchecked energy consumption of AI data centers. On March 26, 2026, Senators Josh Hawley (R-MO) and Elizabeth Warren (D-MA) sent a formal letter to the Energy Information Administration demanding mandatory annual reporting of data center power usage – the latest move in a rapidly escalating campaign that now includes at least four federal bills, state-level moratoriums, and civil penalties of up to $1 million per day for non-compliant facilities.

The legislative push comes as data center electricity consumption in the United States has surged past 25 gigawatts, driven primarily by the artificial intelligence boom that has transformed companies like Google, Microsoft, Amazon, and Meta into some of the largest electricity consumers in the country. With data center demand projected to nearly triple by 2035, lawmakers are racing to prevent what they describe as a looming crisis for residential ratepayers and grid reliability.

This is no longer a niche energy policy debate. The collision between Big Tech’s insatiable appetite for compute power and the physical limits of the American electrical grid has become one of the defining technology policy battles of 2026, with implications that stretch from Wall Street valuations to household electricity bills.

The Hawley-Warren Letter: A Bipartisan Warning Shot

The March 26 letter from Senators Hawley and Warren to the Energy Information Administration represents a significant escalation in Congressional oversight of data center energy consumption. The bipartisan pair demanded that the EIA implement mandatory annual electricity reporting requirements for data centers, including granular data on hourly, annual, and peak energy loads.

Specifically, the senators requested disclosures covering electricity rates paid by operators, grid upgrades needed for new data center loads, how those upgrades are funded, and whether facilities participate in demand response programs. The letter highlights a fundamental gap in federal energy tracking: despite data centers consuming an increasingly massive share of U.S. electricity, no standardized federal reporting mechanism exists to measure or monitor their impact on the grid.

“We have spent years watching data center demand grow exponentially while flying blind on the actual numbers,” Senator Warren stated in the accompanying press release. “Residential ratepayers deserve to know how much of their electricity bill is subsidizing AI training runs for trillion-dollar corporations.”

The letter also referenced Google’s own disclosures showing that the company’s data centers doubled their electricity consumption between 2020 and 2024 – a trend that has only accelerated with the rollout of AI services like Gemini and the expansion of cloud computing infrastructure. The senators argued that without mandatory reporting, grid operators cannot adequately plan for the capacity additions needed to prevent blackouts and price spikes.

The GRID Act: $1 Million Daily Penalties and Mandatory Off-Grid Power

The most aggressive piece of federal data center energy regulation is the Guaranteeing Rate Insulation from Data Centers Act, or GRID Act (S. 3852), introduced on February 12, 2026, by Senators Richard Blumenthal (D-CT) and Josh Hawley (R-MO). The bill targets new data centers with power demand of 20 megawatts or more – essentially every hyperscale facility operated by major cloud providers.

The GRID Act’s core requirement is straightforward but transformative: new qualifying data centers must derive all energy, including backup power, from on-site power generation, captive power plants, or other non-grid sources. Existing facilities receive a 10-year transition window to comply. Facilities that draw power from the grid in violation of the law face civil penalties of up to $1 million per day.

The bill also establishes a “zero rate effect credits” system. Existing data centers that remain connected to the grid must pay credits to offset any increase in residential electricity rates attributable to their demand. The Department of Energy would conduct annual studies to determine whether each facility has contributed to rate increases and calculate the credits owed. Facilities cleared by the DOE receive a Zero Rate Effect Certificate.

“The artificial intelligence revolution should not come at the expense of American families keeping the lights on,” Senator Blumenthal said during a February press conference. “If these companies can raise $100 billion in a single funding round, they can build their own power plants.”

Industry critics have pushed back hard. The Center for Data Innovation called the GRID Act “the wrong way to protect consumers from price spikes,” arguing that forcing data centers off-grid would slow AI development, increase costs, and potentially push investment overseas. Technology trade groups estimate that compliance with the GRID Act’s off-grid mandate could add $500 million to $2 billion in upfront costs per hyperscale facility, depending on the power generation technology chosen.

Four Federal Bills Targeting Data Center Energy: A Legislative Breakdown

The GRID Act is not operating in isolation. At least four major federal bills have been introduced in 2026 targeting data center energy consumption, each taking a different approach to the same fundamental problem. The legislative landscape reflects the depth of concern – and the range of proposed solutions – across the political spectrum.

Bill Name	Lead Sponsor(s)	Date Introduced	Key Provision	Penalty/Enforcement
GRID Act (S. 3852)	Blumenthal (D-CT), Hawley (R-MO)	February 12, 2026	New data centers (20MW+) must use off-grid power	$1 million/day civil penalty
Power for the People Act (S. 3682)	Van Hollen (D-MD)	January 15, 2026	Data centers pay for grid capacity expansion costs	Cost allocation enforcement via FERC
DATA Act	Cotton (R-AR)	February 2026	Creates off-grid utility category exempt from FERC	Exemptions revoked if grid connected
Data Center Water and Energy Disclosure Act	Durbin (D-IL)	March 26, 2026	Mandatory energy and water consumption transparency	Reporting requirements via EIA

Senator Chris Van Hollen’s Power for the People Act takes a different approach than the GRID Act. Rather than forcing data centers off the grid entirely, it requires them to pay for the full cost of grid capacity expansions their demand triggers, preventing those costs from being passed to residential and small business customers. Van Hollen has been actively seeking a Republican co-sponsor to strengthen the bill’s bipartisan credentials.

Senator Tom Cotton’s DATA Act offers what some analysts view as the most industry-friendly approach. It creates a new regulatory category called “consumer-regulated electric utilities” (CREUs) for fully off-grid power systems serving data centers. CREUs would be exempt from FERC rate regulation, reliability standards, interconnection queue processes, and transmission planning requirements. Proponents argue this could reduce project timelines by four to six years by bypassing the interconnection queue – a bottleneck that currently delays data center projects across the country.

Senator Dick Durbin’s Data Center Water and Energy Disclosure Act, introduced on March 26, 2026, focuses specifically on transparency, requiring data centers to report both energy and water consumption data. The bill comes amid growing concern about the water footprint of AI data centers, with some hyperscale facilities consuming millions of gallons per day for cooling systems.

Why Now: The Numbers Behind the Data Center Energy Crisis

The legislative urgency is driven by numbers that would have seemed implausible just five years ago. According to industry estimates, U.S. data centers consumed approximately 25 gigawatts of power capacity in 2025, a figure that has grown by roughly 30% annually since the launch of large-scale AI training workloads in 2023. By 2030, that demand is projected to reach 50 gigawatts or more. By 2035, planned new data center capacity could nearly triple the sector’s total energy footprint.

To put these numbers in context, 25 gigawatts is roughly equivalent to the electricity consumption of the entire state of New York. The projected 2035 demand would exceed the combined electricity consumption of New York, New Jersey, and Connecticut. This is electricity that must come from somewhere – either from new generation capacity that takes years to build or from existing supply that currently serves homes, businesses, hospitals, and schools.

“The math simply doesn’t work,” said Dr. Sarah Chen, an energy systems researcher at MIT. “We cannot simultaneously electrify transportation, decarbonize heating, and power an exponentially growing fleet of AI data centers without massive new generation investment. Something has to give, and right now it’s residential ratepayers.”

The Federal Energy Regulatory Commission has already begun responding to grid stress. In December 2025, FERC directed PJM – the nation’s largest grid operator, serving 65 million people across 13 states – to establish new transparent rules for data center interconnection and large load management. The directive acknowledged that the existing interconnection queue, already backlogged with years of pending applications, is fundamentally unprepared for the scale of data center demand in the pipeline.

Virginia’s Loudoun County, which hosts the highest concentration of data centers in the world, has become the poster child for the tensions. Local officials have reported that data center electricity demand has contributed to measurable increases in residential electricity rates, even as the facilities generate relatively few permanent jobs compared to their energy footprint. Similar concerns have emerged in Georgia, Texas, and Arizona, where major data center expansions are underway.

State-Level Moratoriums Signal Growing Backlash

While federal legislation moves through committee, several states have taken more immediate action – in some cases, halting data center construction entirely. The state-level response underscores just how rapidly the political dynamics around data center energy regulation have shifted, particularly in communities directly affected by the AI data center power crisis.

New York has introduced legislation (A 10141/S9144) that would impose a moratorium on all data center construction for up to three years while state agencies develop rules to minimize impacts on utility rates and grid reliability. South Dakota’s SB 232 proposes a one-year moratorium on hyperscale data center expansion. Oklahoma’s SB 1488 targets data centers with electrical loads exceeding 100 megawatts, imposing a construction freeze until November 2029 pending a thorough study of impacts on water supply, utility rates, and property values.

Arizona and Washington have taken different approaches. Arizona’s HB 2756 would ensure that grid connection costs for data centers cannot be shifted to other retail customers. Washington’s proposed legislation (HB 2515/SB 6171) would bar large data centers from participating in state environmental incentive and emissions credit programs – a move that could significantly increase operating costs for facilities that have relied on green energy credits to meet corporate sustainability targets.

“What we’re seeing at the state level is a rejection of the premise that data centers are inherently good for local economies,” said Mark Thompson, a utilities regulation analyst at Wood Mackenzie. “These facilities bring massive infrastructure costs, relatively few jobs per megawatt consumed, and increasingly, measurable rate increases for residents. The political calculus has fundamentally changed.”

The Trump Administration’s Balancing Act

The federal legislative push exists in tension with the Trump administration’s broader AI strategy, which has emphasized rapid data center buildout as essential to maintaining American technological dominance. Earlier in March 2026, the White House convened a meeting with Big Tech executives to discuss a voluntary compact designed to ensure that data centers do not raise household electricity prices or undermine grid reliability.

The administration’s approach reflects a preference for voluntary industry commitments over mandatory regulation – a stance that has drawn criticism from legislators on both sides of the aisle. Senator Hawley, despite his Republican affiliation, has been publicly skeptical of voluntary measures, arguing that Big Tech companies have repeatedly demonstrated an unwillingness to self-regulate on issues that affect their bottom line.

The SoftBank-funded Stargate project, which envisions up to $500 billion in AI infrastructure investment, illustrates the scale of the tension. The administration has championed the project as a jobs and innovation engine, while critics point out that the electricity demands of Stargate-class data centers could require the equivalent of multiple new power plants – costs that may ultimately flow to ratepayers.

On March 25, 2026, Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez announced plans to introduce legislation that would halt all new data center construction until thorough AI regulations are established. While the bill is unlikely to pass in its current form, it signals the most aggressive position yet in the Congressional debate and may shift the Overton window for what constitutes a “moderate” regulatory approach.

Industry Impact: Big Tech’s $700 Billion Infrastructure Bet at Risk

The legislative activity has sent ripples through the technology sector, where companies have committed hundreds of billions of dollars to AI infrastructure expansion. Big Tech’s combined $700 billion AI infrastructure spending plans were premised on the assumption that electricity would remain abundant and relatively inexpensive – an assumption that now faces regulatory challenge.

Microsoft, which has committed approximately $150 billion to AI capital expenditure through 2026, has been among the most proactive in securing alternative power sources. The company has signed agreements for nuclear power, including a controversial deal to restart the Three Mile Island nuclear plant, and has invested in small modular reactor technology. Amazon has similarly pursued nuclear and renewable energy agreements to power its AWS data centers.

Google, whose data centers were specifically cited in the Hawley-Warren letter for doubling consumption between 2020 and 2024, faces particular scrutiny. The company’s 2025 environmental report acknowledged that its AI workloads had significantly increased energy consumption beyond its prior sustainability projections, and it has struggled to meet its net-zero targets as AI demand outpaces renewable energy procurement.

Company	Estimated 2025 Data Center Power (GW)	2026 AI Capex (Billions)	Power Strategy	Regulatory Exposure
Microsoft	~5.5	$80+	Nuclear (TMI restart), SMRs, renewables	High (Azure expansion in Virginia, Texas)
Amazon (AWS)	~5.0	$75+	Nuclear PPAs, solar farms, natural gas	High (Northern Virginia, Oregon clusters)
Google	~4.5	$60+	Geothermal, renewables, grid PPAs	Very High (cited in Senate letter)
Meta	~4.0	$55+	Solar, wind, Nebius partnership	Moderate (distributed footprint)
Oracle	~2.0	$15+	Natural gas, co-location partnerships	Moderate (rapid expansion phase)

“The GRID Act, if passed in anything close to its current form, would fundamentally alter the economics of hyperscale data center construction,” said David Reynolds, a senior infrastructure analyst at Goldman Sachs. “An additional $500 million to $2 billion per facility for off-grid power generation would compress margins and potentially delay projects that are already in the pipeline.”

The Interconnection Queue Bottleneck

One of the less visible but most consequential dimensions of the data center energy regulation debate is the interconnection queue – the process by which new electricity consumers and generators connect to the grid. Across the United States, interconnection queues are backlogged by years, with wait times of three to five years now common for large load applications. Data centers, which require enormous amounts of power delivered with exceptional reliability, have been particularly affected.

The Cotton DATA Act attempts to address this bottleneck by creating a regulatory bypass for off-grid facilities. By classifying fully self-powered data centers as “consumer-regulated electric utilities” exempt from FERC oversight, the bill would allow operators to skip the interconnection queue entirely – potentially saving four to six years in project development timelines. New Hampshire passed similar legislation in 2025, providing a state-level precedent.

However, critics argue that bypassing FERC oversight creates its own risks. Exempt facilities would not be subject to grid reliability standards, transmission planning requirements, or merger approval processes. If a large off-grid facility later needed to connect to the grid during an emergency or equipment failure, the absence of pre-established interconnection infrastructure could create safety and reliability hazards.

“You can’t have it both ways,” said Jennifer Martinez, a former FERC commissioner now advising state utility commissions. “Either you’re part of the grid and subject to its rules, or you’re entirely self-sufficient — including during emergencies. The DATA Act needs to address the gray area between those extremes.”

Water Consumption: The Hidden Cost of AI Data Centers

Senator Durbin’s Data Center Water and Energy Disclosure Act highlights an increasingly urgent concern that extends beyond electricity: water consumption. Modern data centers require enormous volumes of water for cooling, with a single hyperscale facility consuming between 1 million and 5 million gallons per day, depending on climate, cooling technology, and workload intensity.

AI workloads are particularly water-intensive. A 2025 study by researchers at the University of California, Riverside estimated that training a large language model like GPT-4 consumed approximately 700,000 liters of fresh water for cooling alone. Inference workloads – the ongoing process of running AI models to serve user queries – collectively consume far more water than training, given their continuous operation across thousands of servers.

The water issue has become especially politically sensitive in water-stressed regions. Arizona, which is experiencing chronic drought conditions, has seen significant opposition to new data center projects in the Phoenix metropolitan area. Oklahoma’s moratorium legislation explicitly includes water supply impact assessments alongside energy considerations.

Google disclosed in its 2025 environmental report that its global data center water consumption increased by 20% year-over-year, driven largely by AI workloads. Microsoft has committed to becoming “water positive” by 2030 but acknowledged that AI-driven growth has made that target more challenging. The lack of mandatory federal disclosure requirements for water consumption means that thorough industry-wide data remains unavailable, which is precisely what Durbin’s bill aims to address.

Historical Context: From Dot-Com to AI Boom

The current data center energy regulation debate has historical parallels that illuminate both the opportunities and risks of legislative intervention. During the dot-com era of the late 1990s, the rapid buildout of internet infrastructure triggered similar concerns about electricity demand, particularly in Silicon Valley and Northern Virginia. However, the scale of today’s AI-driven demand dwarfs anything seen during the internet’s early expansion.

In 2000, data centers in the United States consumed approximately 12 terawatt-hours of electricity annually. By 2020, that figure had grown to roughly 73 terawatt-hours. The AI boom has accelerated the trajectory dramatically – industry estimates for 2025 range from 130 to 150 terawatt-hours, with projections for 2030 exceeding 300 terawatt-hours under aggressive AI adoption scenarios.

The regulatory response has also evolved. In the early 2000s, states competed aggressively to attract data centers through tax incentives, subsidized electricity rates, and streamlined permitting. Virginia’s data center tax incentive program, launched in 2009, helped transform Loudoun County into the world’s largest data center market. Those same incentive structures are now under scrutiny as communities question whether the economic benefits justify the infrastructure costs.

The key difference between then and now is AI. Traditional cloud computing workloads were relatively predictable in their power consumption growth. AI training runs, by contrast, can consume orders of magnitude more electricity per unit of compute, and the demand curve shows no signs of flattening. Each successive generation of AI models requires substantially more training compute than its predecessor, creating an exponential growth dynamic that grid planners have struggled to model accurately.

Expert Predictions: What Happens Next

The data center energy regulation landscape is evolving rapidly, and industry analysts, energy experts, and policy observers are offering a range of predictions about what comes next.

Prediction 1: Mandatory federal reporting by 2027. The Hawley-Warren letter and Durbin’s disclosure bill represent the lowest-friction regulatory path. Energy policy experts widely expect some form of mandatory data center energy reporting to be enacted within 12 to 18 months, regardless of the fate of more aggressive bills like the GRID Act. The EIA has the administrative infrastructure to implement reporting requirements relatively quickly once mandated.

Prediction 2: The GRID Act’s off-grid mandate will be softened. While the $1 million per day penalty has generated headlines, most legislative observers expect the GRID Act to undergo significant amendment before reaching a floor vote. The most likely outcome is a hybrid approach that allows grid connection with mandatory cost-sharing provisions, rather than a complete off-grid requirement. “The off-grid mandate is a negotiating position, not a final law,” said Robert Kim, energy policy director at the Bipartisan Policy Center.

Prediction 3: Nuclear power becomes the default solution for hyperscale AI. The convergence of data center energy regulation with growing acceptance of nuclear power – including small modular reactors – positions nuclear as the most likely long-term power source for hyperscale AI data centers. Microsoft’s Three Mile Island restart, Amazon’s nuclear PPAs, and Google’s geothermal investments all point in this direction. By 2030, new data center proposals without dedicated clean energy generation may face significant permitting resistance.

Prediction 4: At least three states will impose construction moratoriums by year-end 2026. The legislative activity in New York, South Dakota, Oklahoma, and other states suggests that data center moratoriums will become more common before federal legislation provides a unified framework. States with acute water stress or grid reliability concerns are the most likely candidates.

Prediction 5: Data center energy costs will become a material factor in AI pricing. If any significant portion of the proposed data center energy regulation is enacted, the increased infrastructure and compliance costs will flow through to AI service pricing. This could accelerate the shift toward more energy-efficient AI architectures and increase demand for inference optimization technologies that reduce per-query energy consumption.

The Competitive Landscape: Global Implications

U.S. data center energy regulation does not exist in a vacuum. The global race for AI infrastructure dominance means that regulatory decisions in Washington have immediate implications for investment flows across borders. China, the European Union, and the Gulf states are all aggressively building data center capacity, and any regulatory friction in the United States could redirect investment to jurisdictions with fewer constraints.

The European Union has taken a different regulatory approach through the Energy Efficiency Directive, which requires data centers above 500 kilowatts to report energy performance indicators beginning in 2024. The EU framework emphasizes transparency and efficiency standards rather than the more aggressive off-grid mandates proposed in the United States. Singapore, which imposed a data center moratorium from 2019 to 2022, has since reopened applications with strict energy efficiency requirements.

China’s approach has been to integrate data center planning directly into national energy strategy, locating new AI compute capacity in regions with surplus renewable energy – particularly in Inner Mongolia, Guizhou, and Ningxia provinces. The “Eastern Data, Western Computation” initiative aims to match data center demand with available energy supply, an approach that some U.S. policy analysts have cited as a more coordinated alternative to the fragmented federal-state regulatory landscape emerging in America.

“The risk is that heavy-handed data center energy regulation pushes investment overseas while doing nothing to reduce global AI energy consumption,” warned Lisa Park, chief technology officer at a major cloud infrastructure provider. “AI training will happen somewhere. The question is whether it happens in the United States under environmental standards, or in jurisdictions with fewer protections.”

What This Means for the Cloud Computing Industry

The immediate impact of data center energy regulation proposals has been felt across the cloud computing supply chain. NVIDIA’s data center connectivity investments and Microsoft’s AI infrastructure spending must now account for regulatory uncertainty as a material risk factor. Cloud providers that have locked in long-term power purchase agreements are better positioned than those still dependent on grid power in regions with active legislative proposals.

For enterprise cloud customers, the regulatory landscape introduces a new variable into procurement decisions. Organizations evaluating cloud providers may increasingly factor energy strategy and regulatory resilience into their selection criteria. Cloud regions located in states with data center moratoriums or aggressive energy regulation may face capacity constraints, potentially affecting service availability and pricing.

The colocation market is particularly exposed. Unlike hyperscale operators with the financial resources to build off-grid power infrastructure, colocation providers typically rely on grid power and pass electricity costs through to tenants. The GRID Act’s 20-megawatt threshold could capture many colocation facilities, forcing a fundamental restructuring of their business model.

Data center REITs have already begun pricing in regulatory risk. Equinix, Digital Realty, and CyrusOne have all referenced data center energy regulation in their Q1 2026 earnings guidance, noting potential impacts on development timelines and capital costs. The uncertainty itself is costly – delayed permitting decisions and shifting regulatory requirements can add months or years to project schedules, increasing carrying costs and reducing returns.

The Path Forward: Compromise or Confrontation

The top shape of data center energy regulation will likely be determined by whether Congress can find a bipartisan compromise that balances energy affordability, grid reliability, and AI competitiveness. The existing legislative proposals span the full spectrum – from Cotton’s industry-friendly DATA Act to the Sanders-Ocasio-Cortez construction moratorium – suggesting that the final outcome will involve significant negotiation.

The most likely near-term outcome is passage of transparency and disclosure requirements, which command the broadest bipartisan support and face the least industry opposition. Mandatory EIA reporting of data center energy consumption would at least provide the data foundation needed for evidence-based policy, addressing a gap that both industry advocates and regulators acknowledge.

The more contentious questions – whether data centers should be required to self-generate power, how infrastructure costs should be allocated between operators and ratepayers, and whether moratoriums are justified – will likely play out over a longer legislative timeline. The 2026 midterm elections could shift the political dynamics significantly, particularly if energy costs become a salient voter issue in states with large data center concentrations.

What is already clear is that the era of unregulated data center energy consumption is ending. The bipartisan nature of the legislative push – with conservative Republicans like Hawley and Cotton joining progressive Democrats like Warren and Blumenthal – suggests that some form of data center energy regulation is inevitable. The remaining question is not whether regulation will come, but how aggressive it will be and how quickly it will reshape the economics of AI infrastructure.

Related Coverage

• The AI Data Center Power Crisis: How Big Tech’s 125 GW Appetite Is Reshaping the US Energy Grid
• Big Tech’s $700 Billion AI Infrastructure Bet: Inside the 2026 Spending Race
• Microsoft’s $150 Billion AI Capex Gamble: Inside the Azure Surge and Copilot’s Adoption Crisis
• NVIDIA’s $4 Billion Silicon Photonics Bet: Inside the Lumentum and Coherent Deals Reshaping AI Data Center Connectivity
• SoftBank’s $40 Billion Loan for OpenAI: Inside the Largest AI-Focused Corporate Loan in History
• Meta’s $27 Billion Nebius Deal: Inside the AI Infrastructure Agreement Reshaping the Data Center Market

Frequently Asked Questions

What is the GRID Act and how does it affect data centers?

The Guaranteeing Rate Insulation from Data Centers Act (GRID Act), introduced by Senators Blumenthal and Hawley in February 2026, requires new data centers with power demand of 20 megawatts or more to generate their own electricity from off-grid sources. Existing facilities have 10 years to comply. Non-compliant facilities face civil penalties of up to $1 million per day. The bill aims to prevent data center energy demand from increasing residential electricity rates.

How much electricity do US data centers consume?

U.S. data centers consumed approximately 25 gigawatts of power capacity in 2025, equivalent to roughly 130 to 150 terawatt-hours annually. This is approximately equal to the total electricity consumption of New York state. Demand is projected to double by 2030 and nearly triple by 2035, driven primarily by AI workloads.

Which senators are leading data center energy regulation?

The effort is bipartisan. Key senators include Josh Hawley (R-MO), Elizabeth Warren (D-MA), Richard Blumenthal (D-CT), Chris Van Hollen (D-MD), Tom Cotton (R-AR), and Dick Durbin (D-IL). Senators Bernie Sanders and Alexandria Ocasio-Cortez have proposed the most aggressive measure – a construction moratorium until AI regulations are established.

Will data center energy regulation increase cloud computing costs?

Industry analysts estimate that the GRID Act’s off-grid power mandate could add $500 million to $2 billion in upfront costs per hyperscale facility. These costs would likely flow through to cloud service pricing over time. Even less aggressive regulations requiring cost-sharing or transparency reporting could introduce new compliance expenses that affect cloud economics.

What states have imposed data center construction moratoriums?

As of March 2026, several states have introduced moratorium legislation. New York’s proposed bill would halt construction for up to three years. South Dakota has proposed a one-year moratorium on hyperscale expansion. Oklahoma has proposed freezing construction of data centers exceeding 100 megawatts until 2029. Arizona and Washington have introduced less restrictive measures targeting cost allocation and environmental incentives.

How does data center energy regulation in the US compare to other countries?

The European Union requires data centers above 500 kilowatts to report energy performance indicators. Singapore imposed a moratorium from 2019 to 2022 before reopening with efficiency requirements. China integrates data center planning into national energy strategy through its “Eastern Data, Western Computation” initiative. The U.S. approach is currently fragmented across multiple competing federal bills and state-level actions.

What role does water consumption play in data center regulation?

Water consumption has emerged as a major concern alongside electricity. A single hyperscale data center can consume between 1 million and 5 million gallons of water per day for cooling. AI workloads are particularly water-intensive – training a large language model can consume approximately 700,000 liters of fresh water. Senator Durbin’s Data Center Water and Energy Disclosure Act would mandate reporting of both energy and water consumption.