Bernie Sanders is right to take AI data centers seriously. But the real question is bigger than Bernie: if AI is a revolution, who gives it direction?
A pause is not always panic.
Sometimes a pause is the first serious sign that the public has noticed what is being built around it.
That is how the AI data center moratorium debate should be understood.
The point is not that artificial intelligence should be stopped.
The point is that AI is no longer only a software question. It is becoming a physical infrastructure question.
Electricity.
Water.
Land.
Labor.
Local communities.
Public subsidies.
Grid capacity.
Electric bills.
And the right of the public to ask what kind of future is being built with shared resources.
That was the core argument of the first article in this series: AI data centers are not just buildings behind fences. They are claims on public systems. They draw from the grid, from land, from water, from local planning decisions, from labor markets, and from political permission.
This is why Bernie Sanders and Alexandria Ocasio-Cortez matter in this debate.
Not because they have solved AI governance.
They have not.
Not because one senator and one representative can direct a planetary technological revolution.
They cannot.
They matter because they are treating AI infrastructure as infrastructure.
That sounds obvious, but it is not how the public conversation usually works.
Too much of the AI debate still sounds like it is about chatbots, model benchmarks, productivity tools, office work, cheating on homework, or vague future risks.
But the concrete question is already here:
Who gets to build the physical infrastructure of the AI age?
Who pays for it?
Who benefits from it?
Who absorbs the risk?
And who gets asked before the concrete is poured?
The Sanders/Ocasio-Cortez AI Data Center Moratorium Act is not important because every part of it is perfect. It is important because it breaks the spell of inevitability.
It says that AI data center buildout should not simply race ahead until national safeguards exist around safety, workers, public benefit, electricity prices, environmental harm, community approval, public subsidies, and labor standards.
The proposal is also more specific than a casual reading may suggest.
It is not simply a ban on every server room, every small facility, or every ordinary data center. In the section-by-section material, the target is large AI-related infrastructure: facilities used for AI model development or deployment at scale, including high-power sites above a 20 MW peak-load threshold or facilities built around high-performance racks and liquid cooling.
That precision matters.
It makes the argument harder to dismiss as technophobia.
The question is not whether every computer rack should be treated as a national emergency.
The question is whether large-scale AI infrastructure should be allowed to lock in major claims on public systems before the public has rules, visibility, and bargaining power.
That is a serious political signal.
But it is still only a signal.
A U.S. moratorium cannot govern a global AI infrastructure race by itself. Compute can move. Capital can move. Workloads can move. Models can be trained in one jurisdiction, deployed in another, financed in a third, and used everywhere.
So Bernie is not the answer.
Bernie is evidence that the question has become serious.
And the question is this:
If AI is revolutionary, then what is the revolution for?
Because “revolution” is not a moral category.
A revolution does not automatically mean liberation.
It means structures change quickly.
It means power moves.
It means old systems become unstable.
It means new actors gain leverage.
That can be good.
It can also be capture at historical speed.
Right now, most of the direction is being set by default.
By investment cycles.
By chip supply.
By land deals.
By cloud contracts.
By grid queues.
By national security pressure.
By corporate competition.
By whoever can build before anyone can say no.
That is not public direction.
That is acceleration filling the space where governance should be.
This is also why the direction of AI cannot be outsourced to the builders alone.
Parts of the AI industry have taken these risks seriously. It would be lazy to pretend otherwise.
OpenAI’s charter says AGI should benefit all of humanity and should not be used in ways that harm humanity or unduly concentrate power.
Anthropic describes itself as an AI safety and research company working to build reliable, interpretable, and steerable AI systems.
The builders were not silent.
Some warned early.
Some warned seriously.
Some likely understood the scale of the problem before many governments did.
That deserves respect.
But warnings are not governance.
A founder can sincerely warn about risk and still operate inside a market race.
A lab can publish safety frameworks and still need compute, capital, talent, distribution, and speed.
A company can speak about humanity and still be structurally pressured to win.
That does not automatically make the company evil.
It means the direction of a revolution cannot depend on the conscience of its builders alone.
The OpenAI/Anthropic divide should be read in that light.
It should not be turned into a children’s story where one side is morally pure and the other side is morally corrupt.
The more important point is simpler: even inside frontier AI, the moral and institutional direction of the technology was contested.
That should tell us something.
If even the builders disagreed about the moral direction of the revolution, the public cannot outsource that direction to the builders.
There is even a slightly funny cultural marker here.
When Sam Altman appears in an Oprah special about “AI and the Future of Us,” AI has left the lab. It has entered mass culture. It has entered the living room. It has become something ordinary people are expected to understand, fear, use, and adapt to.
That is not an energy source.
It is not a technical source.
It does not prove anything about megawatts, water use, or permitting.
But it marks the cultural moment clearly:
AI became mainstream before democratic direction caught up.
That gap is the issue.
The problem is not that nobody warned.
The problem is that warning did not create governance.
The problem is not that nobody saw risk.
The problem is that seeing risk did not decide direction.
And the physical case is now too large to ignore.
The strongest argument for taking AI data centers seriously is not abstract fear.
It is power.
U.S. data centers consumed about 4.4 percent of total U.S. electricity in 2023 and are projected to consume roughly 6.7 to 12 percent by 2028, depending on growth assumptions. Total U.S. data center electricity use rose from 58 TWh in 2014 to 176 TWh in 2023, with projections between 325 and 580 TWh by 2028.
That is the hard center of the argument.
Not because electricity numbers explain everything.
They do not.
But because they turn AI from a speculative debate into a physical planning problem.
Once AI infrastructure becomes large enough to affect grid planning, power prices, local siting, water systems, permitting, labor standards, public subsidies, and community approval, it no longer belongs only to the technology industry.
It belongs to infrastructure politics.
Globally, the International Energy Agency projects that data center electricity consumption could more than double to around 945 TWh by 2030.
That number is significant.
But it must be read honestly.
The same IEA analysis says that 945 TWh would still represent just under 3 percent of total global electricity consumption in 2030, and that the global share remains limited.
A serious argument should not pretend that global electricity demand will be swallowed by data centers.
That is not the claim.
The stronger claim is more specific:
National load growth.
Local grid stress.
Electricity prices.
Water use.
Land use.
Permitting.
Public subsidies.
Community bargaining power.
The world may not be swallowed by data centers.
But a specific grid can still be strained.
A specific town can still be reshaped.
A specific water system can still come under pressure.
A specific household can still face higher bills.
A specific community can still be asked to absorb public risk for private acceleration.
The issue is not only how large AI infrastructure becomes in the global total. It is where it lands, who pays for the buildout, who gets priority on the grid, who receives the benefits, and whether local communities can see and challenge the terms before the infrastructure becomes locked in.
The question is not whether intelligence should exist.
The question is whether public systems should be consumed as private inputs before the public can read the bargain.
This is where public direction must become more than a slogan.
It has to become a mechanism.
AI infrastructure should be visible.
Communities should know who is building, who owns the facility, what power and water it requires, what subsidies it receives, what emissions it creates, what labor standards apply, and what agreements have been made with utilities and public authorities.
AI infrastructure should be verifiable.
The claims made by companies and governments should be backed by public reporting, independent audit, and measurable obligations, not only voluntary promises or press releases.
AI infrastructure should be separable.
Energy supply, public subsidies, land use, water use, labor standards, model deployment, and community approval should not be bundled into one irreversible private deal that the public can only accept or reject after the fact.
AI infrastructure should be replaceable.
If one buildout model creates too much public risk, there must be alternatives: different siting, different power arrangements, different ownership models, different public-benefit requirements, different scale, or no approval.
That is the difference between a directed revolution and a captured one.
A directed revolution remains readable.
A captured revolution becomes infrastructure before society has understood the bargain.
This is why a moratorium matters as a concept, even if the exact policy is incomplete.
It says the future is not only built by whoever pours concrete first.
It says public systems are not passive inputs to private acceleration.
It says electricity, water, land, labor, and local consent are not background conditions.
They are the material constitution of the AI age.
Even outside the United States, the same pattern is appearing.
Europe is moving from voluntary concern toward standards, reporting, and sustainability labels for data centers, including questions around water use and clean energy sourcing.
UN-linked researchers have warned that unchecked data center growth can put combined pressure on energy supplies, water systems, land use, and e-waste if governments do not regulate responsibly.
These are not fringe signals.
They are signs that the AI infrastructure question is becoming public.
But the pieces are still fragmented.
The U.S. has one debate.
Europe has another.
Local communities fight individual sites.
Companies make voluntary commitments.
Researchers publish warnings.
Investors keep funding buildout.
Governments talk about competitiveness.
And the physical infrastructure of the AI age continues to expand.
That is the danger of directionless acceleration.
Not that every project is bad.
Not that every company is corrupt.
Not that every AI benefit is imaginary.
The danger is that the future becomes locked in before the public has defined what the future is for.
So the real choice is not AI or no AI.
That is the wrong frame.
The real choice is directed revolution or captured revolution.
A directed revolution asks:
Who benefits?
Who pays?
Who owns the infrastructure?
Who gets protected?
Who gets displaced?
Who gets a say?
Who gets the upside?
Who bears the downside?
A captured revolution answers those questions quietly, through contracts, capital, speed, and default power.
That is why “anti-AI” is the wrong accusation.
This argument is not against intelligence.
It is against building a new intelligence infrastructure without public direction.
Bernie Sanders is right to take that seriously.
The AI industry deserves credit for warning about real risks.
OpenAI’s moral language matters.
Anthropic’s safety posture matters.
The Oprah moment matters as a cultural marker.
The U.S. energy numbers matter most because they make the argument physical.
The global numbers matter too, but they should be read carefully: the world may not be swallowed by data centers, while specific countries, grids, regions, towns, water systems, and households can still experience serious pressure.
That is exactly why readability matters.
If AI infrastructure is visible, verifiable, separable, and replaceable, then the public can argue about direction before the buildout becomes irreversible.
If it is not, then direction will be decided by default.
By capital.
By speed.
By infrastructure lock-in.
By whoever builds first.
If AI is revolutionary, then the public has the right to ask what direction the revolution is serving.
A revolution without public direction is not liberation.
It is capture at historical speed.
Sources
- Hedegreen Research. "A Moratorium for AI Data Centers." https://hedegreenresearch.com/articles/a-moratorium-for-ai-data-centers/
- U.S. Senator Bernie Sanders. "Sanders, Ocasio-Cortez Announce AI Data Center Moratorium Act." https://www.sanders.senate.gov/press-releases/news-sanders-ocasio-cortez-announce-ai-data-center-moratorium-act/
- U.S. Congress / GovInfo. "S.4214 - Artificial Intelligence Data Center Moratorium Act." https://www.govinfo.gov/content/pkg/BILLS-119s4214is/html/BILLS-119s4214is.htm
- U.S. Senator Bernie Sanders. "Artificial Intelligence Data Center Moratorium Act - Section-by-Section." https://www.sanders.senate.gov/wp-content/uploads/Artificial-Intelligence-Data-Center-Moratorium-Act-Section-by-Section.pdf
- U.S. Department of Energy. "DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers." https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data-centers
- Lawrence Berkeley National Laboratory. "Berkeley Lab Report Evaluates Increase in Electricity Demand from Data Centers." https://newscenter.lbl.gov/2025/01/15/berkeley-lab-report-evaluates-increase-in-electricity-demand-from-data-centers/
- Lawrence Berkeley National Laboratory. "2024 United States Data Center Energy Usage Report." https://eta.lbl.gov/publications/2024-lbnl-data-center-energy-usage-report
- International Energy Agency. "Energy and AI: Energy Demand from AI." https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai
- International Energy Agency. "Energy and AI: Executive Summary." https://www.iea.org/reports/energy-and-ai/executive-summary
- OpenAI. "OpenAI Charter." https://openai.com/charter/
- Anthropic. "Company." https://www.anthropic.com/company
- ABC. "AI and the Future of Us: An Oprah Winfrey Special." https://abc.com/news/1efd942d-61bb-4519-8a62-c4a8fce50792/category/1138628
- Reuters. "EU plans energy standards for data centres amid concerns over soaring power use." 2026-06-03. https://www.reuters.com/business/energy/eu-plans-energy-standards-data-centres-amid-concerns-over-soaring-power-use-2026-06-03/
- Reuters. "AI to double data centre power, water consumption by 2030, UN researchers say." 2026-06-03. https://www.reuters.com/business/energy/ai-double-data-centre-power-water-consumption-by-2030-un-researchers-say-2026-06-03/
