BLACKWIRE

PRISM - TECH & SCIENCE

The Moratorium: Bernie Sanders Wants to Freeze Every AI Data Center in America

A senator from Vermont, a congresswoman from the Bronx, and a growing bipartisan coalition are trying to slam the brakes on the most expensive infrastructure buildout in human history. The AI industry has no good answer for why they shouldn't.

BLACKWIRE • March 30, 2026 • 14 min read

Data centers consume more electricity than many small nations. The political backlash is catching up. (Unsplash)

The bill is five pages long. Its ambitions are enormous. On Wednesday, March 26, Senator Bernie Sanders introduced the Artificial Intelligence Data Center Moratorium Act, a piece of legislation that would halt the construction or upgrading of every AI-focused data center in the United States until Congress passes laws addressing AI safety, environmental harm, rising electricity bills, worker displacement, and the concentration of wealth in the hands of a few tech billionaires. Representative Alexandria Ocasio-Cortez will introduce a companion bill in the House in the coming weeks.

The moratorium has no expiration date. It ends only when Congress acts. Given that Congress has failed to pass meaningful AI legislation since the technology exploded into public consciousness three years ago, that could mean the freeze lasts a very long time.

"A moratorium will give us the chance to figure out how to make sure that AI benefits the working families of this country, not just a handful of billionaires who want more and more wealth and more and more power," Sanders said in a speech on Capitol Hill on Tuesday evening, according to WIRED. "A moratorium will give us the time to figure out how to ensure that AI is safe and effective and prevent the worst outcomes."

The bill will almost certainly not pass. The Trump administration has staked its tech policy on maximum AI acceleration, and the industry is set to spend record amounts on lobbying this year. But dismissing the Sanders bill as political theater misses something important happening beneath the surface. The resistance to data centers is no longer a progressive niche issue. It is becoming one of the rare points where the American left and right agree - and the industry's response so far has been inadequate.

What the Bill Actually Says

The Sanders bill targets any facility with an energy load above 20 megawatts used for AI workloads. (Unsplash)

The Artificial Intelligence Data Center Moratorium Act targets a specific class of facilities. It defines AI data centers through physical parameters, primarily energy load above 20 megawatts. That threshold captures the warehouse-scale computing installations that companies like Microsoft, Amazon, Google, and Meta have been racing to build, while exempting smaller server rooms and traditional hosting operations. According to the bill's section-by-section summary, the moratorium covers both new construction and the upgrading of existing facilities.

The conditions for lifting the moratorium are expansive. Congress would need to enact laws that:

• Prevent data centers from contributing to climate change, harming the environment, and raising electricity bills for consumers
• Prevent tech companies from producing products that harm the health and well-being of working families, privacy and civil rights, and "the future of humanity"
• Ensure that wealth generated from AI is "shared with the people of the United States"

A separate section forbids the export of computing hardware - including semiconductor chips - to any country that does not have similar legislative protections in place. This provision alone, if enacted, would have enormous implications for the global AI supply chain and would effectively force allied nations to adopt Sanders-style AI regulations or lose access to American silicon.

The bill name-checks five wealthy tech executives by name: xAI's Elon Musk, Amazon's Jeff Bezos, OpenAI's Sam Altman, Anthropic's Dario Amodei, and Meta's Mark Zuckerberg. Sanders points to the paradox at the heart of the AI boom - that the same people who profit most from the technology are also the ones sounding the loudest alarms about its dangers. The implication is clear: if these executives genuinely believe AI poses existential risks, they should welcome a pause to get the regulatory framework right.

The Numbers Behind the Backlash

Communities near data centers are reporting double-digit increases in electricity bills. The anger is real. (Unsplash)

Sanders did not arrive at this position in a vacuum. The data center backlash has been building for over a year, and the numbers explain why.

A Pew Research Center survey of 8,512 U.S. adults, conducted January 20-26, 2026, found that Americans view data centers far more negatively than positively across most dimensions. The findings, published March 12, paint a picture of a public that has noticed the buildings going up in their communities and does not like what it sees.

Three-quarters of Americans say they have heard or read at least a little about data centers, with 25 percent saying they have heard a lot. On the environment, 39 percent say data centers are mostly bad versus just 4 percent who say they are mostly good - a nearly ten-to-one ratio. On home energy costs, 38 percent say data centers are mostly bad versus 6 percent who say they are mostly good. On quality of life for nearby residents, the split is 30 percent negative to 6 percent positive.

The only areas where data centers receive more positive than negative ratings are local jobs (25 percent good versus 15 percent bad) and local tax revenue (23 percent good versus 12 percent bad). But even these bright spots come with a caveat: in each category, between 17 and 27 percent of respondents said they were not sure, suggesting that the jury is still out for a significant portion of the public.

Democrats and Democratic-leaning independents view data centers more negatively than Republicans, but the Pew data shows that concerns cross party lines. And the most important metric may be the real-world one: in the second quarter of 2025 alone, $98 billion worth of data center projects were stalled or canceled due to community pushback, according to industry research cited by WIRED. That is not a rounding error. That is a structural brake on the entire AI infrastructure buildout.

The United States now has more than 4,000 data centers, with especially large concentrations in Virginia, Texas, and California. Virginia's Loudoun County - known as "Data Center Alley" - has become ground zero for the backlash, with local elections increasingly turning on candidates' positions on further development. In Georgia, community opposition has reached a level where it is affecting statewide politics. The pattern is repeating across a dozen states.

The Bipartisan Coalition Nobody Expected

The data center moratorium is uniting political factions that agree on almost nothing else. (Unsplash)

The most politically interesting aspect of the data center backlash is who is participating. Sanders and Ocasio-Cortez represent the progressive wing of the Democratic Party. But the voices calling for restrictions on data centers now include some of the most prominent figures on the populist right.

Before Sanders made his first public call for a moratorium in December 2025, Republican Representative Thomas Massie of Kentucky was already questioning the data center buildout on social media. Senator Josh Hawley of Missouri - one of the Senate's most prominent MAGA-aligned members - has been vocally critical. Former Representative Marjorie Taylor Greene raised concerns before leaving Congress. Steve Bannon, one of the most influential anti-AI voices in Washington, hosted a War Room podcast segment in December titled "Data Centers Are Devouring Public Land."

In February 2026, Hawley and Democratic Senator Richard Blumenthal introduced a bipartisan bill specifically designed to insulate consumers from electricity rate hikes caused by data centers. That bill, reported by NBC News, was the first bipartisan legislative effort to address the issue.

Florida Governor Ron DeSantis has been especially vocal. "I don't think there's very many people who want to have higher energy bills just so some chatbot can corrupt some 13-year-old kid online," DeSantis said at an AI roundtable in February, neatly combining energy populism with tech skepticism. He endorsed legislation that would have established a bill of rights protecting consumers from AI harms, prohibited minors from using chatbots without parental consent, stripped subsidies from data center developers, and prohibited the facilities from raising electricity bills. The Senate version passed. The House version died.

At the state level, the movement has exploded. Dozens of cities and counties have introduced local moratoriums on data center development. At least twelve state legislatures - Georgia, Maryland, Minnesota, New Hampshire, New York, Oklahoma, Pennsylvania, South Dakota, Vermont, Virginia, Wisconsin, and Wyoming - have introduced state-level moratoriums in 2026. The bills span the partisan spectrum. Oklahoma's was introduced by Republicans. Georgia's had bipartisan cosponsors. The environmental group Food and Water Watch helped craft the New York bill and advised Sanders's office on the federal version.

"It makes sense to me that his bill is going to focus primarily on that aspect," Mitch Jones, the policy and litigation director at Food and Water Watch, told WIRED, referring to Sanders's broader framing around AI safety rather than environmental concerns alone.

The bipartisan nature of the opposition is a problem for the tech industry that no amount of lobbying money can easily solve. When the left opposes you on environmental and labor grounds and the right opposes you on energy costs and cultural grounds, your traditional playbook of aligning with one party against the other stops working.

Big Tech's Failed PR Offensive

Tech companies spent hundreds of billions building data centers before bothering to explain why communities should accept them. (Unsplash)

The industry knows it has a problem. In March 2026, representatives from top data center developers and AI companies - including Amazon, Microsoft, Meta, and Google - gathered at the White House to sign a nonbinding agreement intended to make data centers pay "the full cost of their energy and infrastructure" and protect consumers from rate hikes.

"Data centers - they need some PR help."

That was President Donald Trump, speaking at the signing event. He was not wrong. But the agreement he hosted was, by most expert assessments, largely symbolic. Experts told WIRED that several of the key aims - including having data centers absorb any additional costs to customers' bills - are largely beyond either the White House's or the tech companies' control. Electricity pricing is primarily regulated at the state level through public utility commissions, and the mechanics of how rate increases get distributed among customer classes involve regulatory proceedings that a White House pledge cannot override.

The Data Center Coalition, an industry group, responded to Sanders's bill with a statement that reads like it was drafted by committee - which it almost certainly was. "A moratorium would limit internet capacity, slow critical services, eliminate hundreds of thousands of high-wage jobs, drain billions in local tax revenue, and raise costs for American families and small businesses," Cy McNeill, the group's senior director of federal affairs, told WIRED.

The statement reveals the industry's core problem. It is still framing data centers as "internet capacity" and "critical services" when the public increasingly sees them as AI factories that benefit a narrow class of technology companies and their investors. The economic arguments about jobs and tax revenue are real but insufficient. Pew's data shows that even in the areas where data centers poll positively - jobs and tax revenue - the margins are modest compared to the overwhelming negative sentiment on environment, energy costs, and quality of life.

The industry is also struggling with a credibility gap of its own making. For years, tech companies told communities that data centers would bring jobs and investment. What many communities got instead were windowless buildings that employ surprisingly few people, consume enormous amounts of water and electricity, and generate noise and heat. A modern hyperscale data center might cost billions to build but employ only a few dozen permanent staff. The construction jobs are temporary. The tax abatements that lured the facilities often expire, leaving communities wondering what they gained.

Microsoft, to its credit, has been more proactive than most. The company has published commitments to carbon negativity and water positivity. But these pledges operate on timelines - 2030, 2050 - that do not help the family in Loudoun County whose electricity bill went up 20 percent last year. The gap between corporate sustainability reports and lived experience is where the political backlash lives.

The Deeper Question: Who Pays for the AI Revolution?

The energy demands of AI training runs are measured in gigawatts. Communities are asking who should foot the bill. (Unsplash)

Strip away the politics and Sanders's bill asks a question that the AI industry has been avoiding: who bears the cost of the AI revolution?

The economics are staggering. Microsoft, Google, Amazon, and Meta collectively announced more than $300 billion in capital expenditure plans for 2026, the vast majority directed at AI infrastructure. Nvidia, which makes the GPUs that power these data centers, reported quarterly revenue exceeding $40 billion. OpenAI is reportedly preparing for an IPO. Anthropic's valuation has climbed past $60 billion. The money flowing into AI infrastructure dwarfs anything the tech industry has previously attempted.

But the costs are not evenly distributed. The profits flow to technology companies, their investors, and their employees - concentrated primarily in the San Francisco Bay Area, Seattle, and a handful of other tech hubs. The costs - higher electricity bills, strained water supplies, environmental degradation, noise pollution, and the aesthetic transformation of rural and suburban landscapes into industrial zones - are distributed across communities that often had no say in the decision to site a data center in their neighborhood.

This asymmetry is the political core of the issue. It is the same dynamic that has fueled opposition to highways, landfills, power plants, and other infrastructure projects throughout American history. The difference is scale. No previous infrastructure buildout has moved this much money this fast with this little regulatory framework in place.

The Sanders bill's demand that AI wealth be "shared with the people of the United States" is vague as policy but clear as politics. It taps into the same populist energy that has driven both left-wing movements like Occupy Wall Street and right-wing movements like the MAGA populist wing. When the richest people in the world are getting richer at an accelerating rate while ordinary people's utility bills are climbing, the political math writes itself.

The chip export provision adds another dimension. By tying hardware exports to the adoption of AI regulations by foreign governments, Sanders is essentially proposing to use America's dominance in semiconductor manufacturing as leverage to force global AI governance. Whether this would work in practice is debatable - China already has its own chip manufacturing ambitions, and allied nations might resent being coerced. But it reflects a growing recognition in Washington that AI governance is inherently a global problem that no single country can solve alone.

The State-Level Laboratory

From Virginia to Georgia to Oklahoma, communities are pushing back against data center sprawl. (Unsplash)

While the federal moratorium faces long odds, the state-level action is where the real regulatory risk lies for the industry. The twelve states that have introduced moratorium bills in 2026 are not outliers. They represent a movement that is still accelerating.

Virginia offers the most instructive case study. The state hosts more data center capacity than anywhere else in the world, concentrated in Loudoun and Prince William counties in the northern part of the state. For years, local officials welcomed the facilities for their tax revenue. Then the bills started arriving.

Dominion Energy, the state's largest utility, has requested rate increases that regulators have linked directly to the infrastructure demands of data centers. Residential customers have seen their bills climb. Water usage has become a contentious issue, with data centers consuming millions of gallons for cooling systems. Traffic from construction has congested rural roads. The hum of generators and cooling systems has become a constant background noise for nearby residents.

The political fallout has been swift. In Virginia's 2025 state elections, data center opposition became a campaign issue in multiple races. Candidates who took anti-data-center positions won in districts that had previously been friendly to the industry. The Virginia legislature has not yet passed a moratorium, but the debate is active and intensifying.

Georgia's experience is similar but more dramatic. The state has become a hotspot for data center development, with companies attracted by relatively low energy costs and available land. But community opposition has been fierce, particularly in rural counties where residents worry about water supplies and the transformation of agricultural landscapes. The bipartisan nature of Georgia's moratorium bill - with both Democratic and Republican cosponsors - reflects the fact that data center opposition does not map neatly onto traditional partisan lines in many communities.

New York's moratorium effort deserves particular attention. Crafted with input from Food and Water Watch, the New York bill focuses on the environmental and energy dimensions. The state's unique energy market - with a partially deregulated system and aggressive renewable energy targets - makes the data center question especially complex. Adding hundreds of megawatts of data center load to a grid that New York is simultaneously trying to decarbonize creates a mathematical tension that no amount of corporate renewable energy credits can resolve.

Oklahoma's bill is notable for different reasons. Introduced by Republicans in a deep-red state, it undermines the industry's narrative that opposition is a progressive phenomenon. Oklahoma lawmakers are responding to the same constituent concerns - water, electricity costs, land use - that drive opposition everywhere. The details differ. The politics converge.

What the Industry Gets Right - and Wrong

The AI industry argues that data centers are as essential as the electrical grid. Communities are not convinced. (Unsplash)

The Data Center Coalition's argument that a moratorium would "limit internet capacity" and "slow critical services" contains a grain of truth buried under a mountain of misdirection. Data centers do support essential services - cloud computing, telemedicine, financial systems, communications infrastructure. A genuine freeze on all data center construction would create real bottlenecks.

But Sanders's bill specifically targets AI-focused facilities, defined by their energy load. Traditional cloud computing and internet infrastructure operate at much lower power densities than AI training clusters. A single Nvidia H100 GPU draws about 700 watts. A training cluster for a frontier AI model might use tens of thousands of these chips simultaneously, pushing total power consumption into the hundreds of megawatts. The distinction between "data center" and "AI data center" is not just semantic. It maps onto a real physical difference in energy consumption.

The industry is also correct that data centers create jobs and tax revenue. But the numbers need context. The construction phase of a large data center can employ thousands of workers, but these are temporary positions. Once operational, a hyperscale facility might employ 30 to 50 permanent workers. Compare that to a manufacturing plant of equivalent investment size, which might employ hundreds or thousands of people on an ongoing basis. The job-to-capital ratio of data centers is among the lowest of any industrial facility type.

Tax revenue is real but often reduced by the very incentive packages that lured the data centers in the first place. Virginia, for example, has offered data center operators hundreds of millions of dollars in tax exemptions over the past decade. The net fiscal impact for host communities is smaller than the gross numbers suggest, and in some cases may be negative when you account for the infrastructure costs - roads, water treatment, grid upgrades - that the facilities demand.

Where the industry gets it most wrong is in its failure to engage honestly with the distributional question. The benefits of AI flow primarily to companies, investors, and knowledge workers. The costs of AI infrastructure fall disproportionately on communities that are often rural, lower-income, and politically underrepresented. This is not a new pattern in American industrial history, but the speed and scale of the AI buildout have compressed a process that usually unfolds over decades into just a few years.

The War Dimension

The Iran conflict has revealed data centers as military targets, adding a national security dimension to the domestic debate. (Unsplash)

There is one dimension of the data center debate that neither Sanders nor the industry has fully reckoned with: the current war in Iran has demonstrated that data centers are legitimate military targets.

Iran has been targeting data centers with both cyber and conventional weapons throughout the conflict, according to AP News, showing how important these facilities have become to the economy, communications, and military information security. The Pentagon views the concentration of computing infrastructure as a potential vulnerability. Adversaries view it as a target set.

This creates a paradox that neither the pro-build nor the anti-build camps have fully addressed. From a national security perspective, the United States needs robust computing infrastructure. But concentrating that infrastructure in a small number of massive facilities creates single points of failure that adversaries can target. The Iran conflict has shown that cyber-physical convergence - attacking digital infrastructure with physical weapons and vice versa - is now standard military doctrine.

Investigators at the Utah-based security firm DigiCert have tracked nearly 5,800 cyberattacks mounted by approximately 50 different groups tied to Iran since the conflict began. While most of these attacks targeted U.S. or Israeli companies, the scope extends to Bahrain, Kuwait, Qatar, and other nations. A pro-Iranian hacking group recently claimed responsibility for infiltrating an account belonging to FBI Director Kash Patel, posting years-old photographs and personal documents.

The national security dimension complicates the moratorium debate in both directions. Hawks will argue that the United States needs more computing capacity, not less, to maintain its military edge. Doves will argue that the concentration of AI infrastructure makes the country more vulnerable to attack. Both are right. The conversation the country needs to have is not about whether to build data centers, but about how to build them - how to distribute them geographically, harden them against attack, ensure they serve national interests rather than just corporate ones, and integrate them into a coherent strategy for technological competition.

Sanders's bill does not engage with this dimension. That is understandable - his focus is on domestic policy and economic justice. But any serious national AI strategy will eventually need to reconcile the military, economic, environmental, and social dimensions of computing infrastructure. Right now, the United States is building that infrastructure according to the logic of private capital - wherever land is cheap, electricity is available, and tax incentives are generous. Whether that is also the right logic from a national security perspective is a question that nobody in Washington is asking loudly enough.

What Happens Next

The moratorium will not pass. But the political forces behind it are not going away. (Unsplash)

The Sanders bill will not become law. Not this year, probably not next year either. The Trump administration is ideologically committed to AI acceleration, and the industry's lobbying apparatus is formidable. Even many Democrats who share Sanders's concerns about AI will hesitate to support a full moratorium, preferring targeted regulation to a blanket freeze.

But the bill serves a purpose beyond its legislative prospects. It establishes a marker. It tells the industry that a serious political figure with a national following is willing to go to the most extreme position on the regulatory spectrum. It shifts the Overton window. Suddenly, proposals that would have seemed radical six months ago - mandatory environmental impact assessments for data centers, requirements that facilities pay the full cost of grid upgrades, community benefit agreements, noise and water usage limits - look moderate by comparison.

The state-level action is more immediately consequential. If even a few of the twelve state moratoriums pass, the industry will face a patchwork of restrictions that complicate planning and increase costs. States that pass moratoriums will effectively redirect data center investment to states that do not - creating a competitive dynamic that could eventually force either federal preemption (the industry's preferred outcome) or a federal framework that sets minimum standards (closer to what the bill's supporters want).

The most likely medium-term outcome is neither a moratorium nor the status quo, but a new regulatory regime that requires data centers to internalize more of their costs. This means paying for grid upgrades rather than having them socialized across all ratepayers. It means environmental review processes similar to those required for other large industrial facilities. It means community benefit agreements that go beyond token gestures. It means transparency about water usage, noise levels, and actual employment numbers.

The AI industry has spent the past three years building as fast as possible, racing to secure computing capacity before competitors do. That race has created genuine value - the capabilities of frontier AI models are real and economically significant. But it has also created a political backlash that is now moving faster than the industry's ability to manage it.

Sanders understands something that many tech executives do not: you can build the most advanced technology in human history, but if the communities that host your infrastructure feel exploited rather than empowered, the politics will eventually catch up with you. They always do.

The moratorium is not the answer. But the questions it asks - who benefits from AI, who pays for it, and who decides - are the right ones. The industry would be wise to start answering them honestly before someone less sympathetic than a Vermont senator writes the rules for them.