Let’s start with the number.

71%.

That is the percentage of Americans who told Gallup they oppose AI data center construction in their area. 48% said they strongly oppose it. Only 7% strongly support it.

This is not an activist poll. This is not a data center resistance organization asking its own members how they feel. This is Gallup, the most credible polling organization in America, asking a representative sample of the American public a straightforward question. And nearly three out of four Americans said no.

That number needs to sit for a moment before we move on.

The AI industry has spent billions of dollars on lobbying, public relations, and political access. It has placed allies in the White House, the Senate, and the regulatory agencies. It has framed its expansion as a matter of national security, economic competitiveness, and American technological supremacy. It has invoked China in every hearing, every press release, and every phone call to every elected official who showed signs of asking questions.

And after all of that, 71% of Americans told Gallup they do not want a data center in their community.

THE RESISTANCE IS WINNING. HERE IS THE PROOF.

Community opposition in the United States alone has blocked or delayed at least $64 billion in data center construction projects since 2024. Globally the number is far larger — $156 billion in projects were stopped or delayed in 2025 alone when all forms of opposition are counted. That number is expected to increase as more projects come before local zoning boards and planning commissions whose members answer to voters rather than to investors.

In Warrenton, Virginia, residents voted out every town council member who supported construction of a local Amazon data center. The new council voted unanimously to approve a zoning change that permanently banned all data centers in the entire town. Not delayed. Not restricted. Banned.

Similar stories have played out in Oregon, Utah, New Jersey, and Massachusetts.

And then came Maine.

This month, Maine became the first state in America to pass a statewide ban on AI data center construction. Not a moratorium. Not a pause pending review. A ban. A state legislature looked at what was happening in communities across the country and decided that the default answer to AI data center proposals in Maine is no.

That is a precedent. When one state legislature passes a statewide ban, the legislators in every other state capitol take notice. They start asking their staff to research what their state’s current rules are. They start getting calls from constituents who read about Maine and want to know why their state has not done the same thing.

The industry understands this dynamic. That is why what happened next is so revealing.

THE NUCLEAR PIVOT

When an industry starts fundamentally rethinking its basic infrastructure model, that is not a sign of confidence. That is a sign of pressure.

A nuclear energy startup called Oklo has proposed combining small modular nuclear reactors directly with AI data centers, creating co-located facilities that generate their own power and can be sited far from existing electrical grids. Far from existing grids means far from municipalities. Far from municipalities means far from zoning boards. Far from zoning boards means far from the 71% of Americans who just told Gallup what they think.

Read that sequence again slowly. The industry’s proposed solution to community opposition is to build its infrastructure somewhere communities cannot reach it.

That is not a technical solution. That is a political one. The industry is not trying to address the concerns that 71% of Americans have about water use, electricity consumption, environmental impact, and utility bills. It is trying to build somewhere those Americans cannot vote.

There are several problems with this plan beyond its fundamental contempt for democratic governance.

Oklo has not built a working reactor. Their first prototype is scheduled to go online late next year and has not been tested at scale. The technology is described even by investors friendly to the company as highly speculative. The timeline from prototype to nationally scaled co-located data center infrastructure is measured in decades, not years.

The communities that are fighting data centers right now do not have decades to wait for the industry to find a workaround. They have zoning board meetings next month.

WHAT THE GALLUP NUMBER ACTUALLY MEANS

71% opposition is not a niche concern. It is not a left-wing concern or a right-wing concern. The Gallup poll found that opposition to AI data center construction was one of the least politically divisive issues in American public life. Republicans oppose them. Democrats oppose them. Rural communities oppose them. Suburban communities oppose them.

The industry’s lobbying apparatus is built for a world where public opinion is divided and persuadable. It is not built for a world where 71% of the public has already made up its mind.

Every elected official in America who represents a community where a data center is proposed now has a Gallup poll telling them that opposing the project is the politically safe position. Approving it is the politically risky one. The electoral math has shifted. The industry’s money is still there. But the votes are moving in the other direction.

THE MAINE PRECEDENT AND WHAT COMES NEXT

Maine’s statewide ban will be challenged in court. The data center industry has the resources to litigate it extensively and the legal arguments are not settled. But the political signal is more important than the legal outcome in the short term.

Maine said no at the state level. The question now is which state says no next.

The communities in Indiana and Virginia and Michigan and Arizona and Tennessee who have been fighting these projects one zoning board at a time now have a model for what state-level action looks like. The organizers who have been building the resistance network for the past three years now have a template to take to their state legislators.

The industry spent billions on lobbying to prevent exactly this moment. The moment when the resistance stops being a collection of local fights and starts being a national political movement with polling numbers, electoral consequences, and state-level legislative victories.

That moment is here.

71%. First state ban. $64 billion blocked.

The cricket is louder than the machine. Gallup just confirmed it.

Sources: Gallup poll 2026 · The Motley Fool May 2026 · Data Center Watch · Community opposition documentation 2024 through 2026

Cricketpocalypse is an independent channel. No corporate funding. No AI company money. Just the facts and the bugs.

The post 71% of Americans Oppose AI Data Centers. The Industry Just Blinked. appeared first on Cricketpocalypse.

This is not an accident. Water is the most politically explosive resource in American and global life. The AI industry understands this. That is why water almost never appears in the public conversation about AI.

This article is about the water. The actual water.

ONE CONVERSATION. ONE WATER BOTTLE.

A 2023 study from the University of California, Riverside estimated that a single conversation with ChatGPT uses approximately 500 milliliters of water. Half a liter. A standard bottle of water. Not metaphorically. Literally. The water is used in the data center cooling systems that prevent the chips generating your answer from overheating.

ChatGPT has over 800 million users. They conduct millions of exchanges daily. The water mathematics are not difficult to do. They are simply never presented to you when you open the app.

Microsoft published data showing that its global water consumption increased by 34% between 2021 and 2022, the period when its AI development was accelerating most rapidly. Google’s water use increased by 20% in the same period. These are billions of gallons of water, drawn from municipal supplies, industrial sources, and in some cases directly from freshwater aquifers, in communities that were not always consulted about whether they had water to spare.

THE CHILE CASE

In 2020, Google received permits to build a $200 million data center in Santiago, Chile. The permit covered the facility’s water needs: 7.6 billion liters annually, drawn from the potable water supply of the Santiago metropolitan area.

Seven point six billion liters. Every year. From drinking water.

Chile is in the grip of its worst drought in recorded history. The country’s water crisis has been building for decades, and has produced water rationing in rural communities, collapsed agricultural livelihoods, and a water rights system that has increasingly concentrated access to water among large industrial users at the expense of small farmers and rural communities.

An environmental activist organization called MOSACAT began protesting the project in 2019. They staged demonstrations that received minimal mainstream coverage. They persisted for four years. They provided technical expert testimony. They filed legal challenges.

In February 2024, Santiago’s Environmental Tribunal suspended the permit. In September 2024, Google announced it would halt the project entirely and start from scratch.

Four years of sustained advocacy. Two billion gallons of annual water extraction stopped. By a small group of activists with no institutional backing, no corporate funding, no political allies. Just documentation, persistence, and a courtroom where the evidence could not be ignored.

This is the most complete victory against a hyperscaler data center anywhere in the world. Most people have never heard of it.

THE AMERICAN WATER WARS

In Tucson, Arizona, the No Desert Data Center Coalition formed in response to a proposal to build a data center campus on 290 acres of living Sonoran Desert, a proposal that would require drawing water from aquifers already under stress. By the time your community finds out about a data center, organizer Lee Ziesche says, they have been working on it for years. Amazon ultimately withdrew from the project after pressure, but the developer has attempted to re-advance it, and the fight continues.

Ireland presents perhaps the most extreme national-scale example. By 2024, data centers consumed 22% of Ireland’s national electricity supply. The grid operator EirGrid imposed what amounts to a moratorium on new data center connections in the Dublin area, effective through at least 2028.

THE CHOICE NOBODY IS TALKING ABOUT

The water consumption of AI data centers is not an inevitable feature of the technology. It is a design choice.

Immersion cooling, which submerges computing hardware directly in a non-conductive liquid, eliminates water evaporation entirely. Air cooling with advanced heat exchange can dramatically reduce water use. Microsoft has built experimental data centers that use zero fresh water for cooling. The technology exists. The engineering is mature.

The reason these alternatives are not universally deployed is cost. The upfront capital cost is higher than building a conventional water-cooled facility, even accounting for the ongoing water expenses.

In other words: the AI industry is choosing cheaper construction over community water security. It is externalizing the cost of its water consumption onto communities because those communities have limited legal mechanisms to force the industry to internalize that cost.

This is not a technology problem. It is a governance problem. And the communities that have successfully pushed back have, in almost every case, done so by forcing the industry to internalize costs it was attempting to externalize.

THE WATER WE OWE THE FUTURE

An aquifer is not a tank. It accumulates over geological timescales. The Ogallala Aquifer, which underlies portions of eight Great Plains states, accumulated its water over millions of years. It is currently being depleted at a rate that exceeds natural recharge by a factor of several hundred. At current depletion rates, significant portions of it will be economically unproductive within decades.

No amount of AI capability will refill it.

The data center that draws from an aquifer today is borrowing water from children who are not yet born. That is not a metaphor. That is a physical description of what aquifer depletion means.

Every community that has fought a data center water permit and won has protected something that cannot be manufactured, cannot be replaced, and cannot be bought back at any price once it is gone.

Sources: University of California Riverside 2023 study · Microsoft sustainability reports · Google environmental reports · Santiago Environmental Tribunal ruling February 2024 · EirGrid capacity statements · Data Center Watch

Cricketpocalypse is an independent channel. No corporate funding. No AI company money. Just the facts and the bugs.

The post Your Water. Their Server. appeared first on Cricketpocalypse.

There is a photograph that changed the world. Taken sometime in the late 1950s, it shows a machine, enormous, industrial, the color of government buildings, standing in a room it almost entirely fills. The machine stores five megabytes of data. Five. The equivalent of a single low-resolution photograph. It weighs over a ton. It requires a team of engineers to maintain.

Now reach into your pocket or your bag or your desk drawer and find a micro SD card. The kind you buy at a gas station for nine dollars. That card stores 128 gigabytes. That is 131,072 megabytes. That is 26,214 times more data than the 1956 machine, stored in something smaller than your thumbnail.

The size difference is not incremental. It is civilizational. It happened within a single human lifetime.

This article is about what happens when you apply that same compression ratio, twenty-five thousand to one, to what the AI industry is building right now, in your state, on your water, with your electricity, through your community’s soil.

THE NUMBER YOU NEED TO KNOW

Forty thousand acres. That is the land footprint being planned for AI data center infrastructure in some of the largest projects currently under construction or proposed in the United States.

To give you a frame of reference: the entire city of Washington DC, the capital of the United States, home to the White House, the Capitol Building, the Lincoln Memorial, and four hundred thousand people, covers approximately 39,000 acres.

We are talking about building something larger than Washington DC. For servers.

THE MATH

Let us do the calculation the AI industry hopes you never do.

Forty thousand acres divided by 25,000 equals 1.6 acres. After one compression generation, roughly ten to twenty years by historical trajectory, the same computing capacity that currently requires a land mass larger than Washington DC requires approximately one and a half acres. A city block with parking.

Divide 1.6 acres by 25,000 again. The answer is 0.000064 acres. Approximately 2.8 square feet. The size of a coffee table. A medium one. Nothing fancy.

This is not science fiction. This is the documented historical trajectory of computing technology applied to its current physical form. The industry knows this. The engineers know this. The executives spending $690 billion in 2026 alone on physical infrastructure know this.

WHAT IT COSTS THE PEOPLE WHO DID NOT DECIDE

The compression calculation tells you what will happen to the technology. It does not tell you what will happen to the land. To the water. To the air. To the electricity bills. To the communities.

Those things do not compress. They do not miniaturize.

An aquifer drained in five years takes ten thousand years to refill. A desert ecosystem paved does not return in any human lifetime. A rural community’s character, its quiet, its darkness, its identity as a place, once industrialized, is changed for generations.

In Tucson, Arizona, a company called Beale Infrastructure, a subsidiary of Blue Owl Capital, reportedly building for Amazon, has proposed a data center campus on 290 acres of living Sonoran Desert. The saguaro cactus, the iconic symbol of the American Southwest, takes seventy-five years to grow its first arm. Individual saguaros live two hundred years or more.

A data center would be built on it. Operated for twenty to thirty years. Then decommissioned when the miniaturization arrives. The saguaro that was there would have been seventy years from its first arm. It will not grow back on any timeline that matters to the humans living nearby.

THE COMPANIES BEHIND THE MASK

One of the most consistent patterns in data center community conflicts is that communities often do not know who they are fighting until the fight is almost over.

In Brandy Station, Virginia, the applicant for a $12 billion data center campus was listed as Culpeper Acquisitions LLC. Only through investigative journalism and public records requests did local advocates discover the likely end user was Google or Meta. In Tucson, Arizona, the applicant was Beale Infrastructure, a private equity subsidiary, with Amazon reportedly as the ultimate end user.

This opacity is deliberate. A permit application from Amazon Web Services for a 290-acre Sonoran Desert data center campus generates a different community response than one from Beale Infrastructure for a technology facility. The anonymity is almost always deliberate. The transparency almost always has to be forced.

THE COMMUNITIES THAT WON

In 2024, the residents of Chesterton, Indiana organized against a proposed $1.3 billion data center campus. They showed up at planning and zoning meetings. They made signs. They called their council members. They testified. The developer withdrew.

On the same day that victory was confirmed, organizer Wendy Reigel received a shipment of 200 No Data Center yard signs she had ordered for the fight. They arrived after the fight was over. She packed every one of them into boxes and shipped them to Peculiar, Missouri, where the next fight was just beginning. The Peculiar city council voted unanimously to reject the project in September 2024.

The total value of data center projects blocked, delayed, or cancelled in 2025 alone: $156 billion.

THE QUESTION THAT REMAINS

In fifty years, the generation of children who inherit the world we are building right now will walk past photographs of today’s AI data centers in whatever museums or digital archives they inhabit. They will see images of forty-thousand-acre campuses, cooling towers, power lines stretching to the horizon, desert land stripped and paved for a generation of computing. They will know that all of it was replaced by something the size of a fingernail.

They will ask: was the damage worth it?

The honest answer, the answer that the $690 billion annual investment is actively preventing anyone from having to say out loud, is that the damage was not inevitable. The communities that bore the cost did not agree to bear it. The water that was taken was not offered freely.

The damage was chosen. By specific people. For specific financial reasons. At the expense of specific communities that had less political power than the companies that made the choice.

That is not progress. That is extraction. And the first step to stopping it is knowing what you are looking at when you see a data center going up in your county.

You are looking at the 1956 refrigerator. It is being built on your water. And it will be obsolete before the saguaro grows its first arm.

Sources: Data Center Watch · Heatmap News · Washington Post · TechPolicy.Press · Visual Capitalist

Cricketpocalypse is an independent channel. No corporate funding. No AI company money. Just the facts and the bugs.

The post They’re Building the 1956 Refrigerator. Again. appeared first on Cricketpocalypse.

Then three of the most powerful men in the AI industry communicated with the administration. Elon Musk. Mark Zuckerberg. David Sacks.

The rule was canceled.

The official reason: we can not let China get ahead.

China had nothing to do with that rule. The rule only applied to American companies, operating in America, deploying systems in America. Canceling it did not slow Chinese AI development by a single second. Canceling it removed a safety review process for systems being deployed in American hospitals and banks, systems that will make decisions about your medical care and your money.

China was not in that phone call. Three American billionaires were.

This is what regulatory capture looks like in the AI era. Not a dramatic scandal. Not a bribe. Not a conspiracy. Just three phone calls, and a safety rule dead before most people knew it existed.

THE LOBBYING NUMBERS

The lobbying numbers make the pattern explicit. By the first quarter of 2026, Anthropic was spending $1.6 million per quarter on federal lobbying. OpenAI was spending $1.5 million. Meta was spending $7.1 million per quarter. The seven largest technology companies spent over $50 million on AI lobbying in nine months of 2025.

In the same period, the entire AI safety research field, all of it, worldwide, received less in grants than OpenAI alone spent on lobbyists.

The companies warning most loudly about AI’s catastrophic dangers are simultaneously spending tens of millions of dollars ensuring that the regulations those warnings imply are never implemented. This is not a contradiction. It is a strategy. The fear generates goodwill. The lobbying ensures that goodwill never becomes binding law.

THE CONFLICT AT THE TOP

David Sacks was appointed by the Trump administration as America’s AI and Crypto Czar, the government official most directly responsible for shaping federal AI policy. He is simultaneously a general partner at Craft Ventures, a venture capital firm with an active portfolio of AI startup investments.

This is not a historical fact. This is the current arrangement. The person setting national AI policy holds financial positions in the industry he is setting policy for. His explicit stated position is that AI regulation should be minimally burdensome, a position that, by a remarkable coincidence, serves the commercial interests of the companies in his portfolio.

He has publicly accused Anthropic of running a sophisticated regulatory capture strategy based on fear-mongering. He is correct about the mechanism. He is also engaged in the inverse version of the same strategy: using a government position to eliminate regulations that would inconvenience his portfolio companies while framing deregulation as American competitiveness against China.

Both Sacks and Anthropic are engaged in regulatory capture. They are capturing in opposite directions. Anthropic wants regulations it helped design. Sacks wants no regulations at all. But both are pursuing outcomes that serve specific financial interests at the expense of democratic governance.

THE CHINA ARGUMENT AND WHY IT DOES NOT HOLD

The single most powerful rhetorical weapon in the AI industry’s regulatory capture arsenal is the China argument. We can not let China beat us, or its variants, appears in virtually every argument against AI safety legislation, environmental review, community consultation, or any other mechanism that might slow American AI development.

First: American AI safety regulations do not affect what China builds. This is a simple empirical fact. A requirement that American AI companies submit to pre-deployment safety review has zero effect on the research output of Chinese AI labs.

Second: The specific regulations being opposed are rarely about AI capability. The executive order that was canceled after phone calls from Musk, Zuckerberg, and Sacks would have required AI companies to show their systems to federal safety experts before releasing them, to protect banks and hospitals. This is not a capability constraint. It is a review process.

Third: The countries that regulate AI most carefully are not losing the competition. The European Union’s AI Act is the most comprehensive AI regulation in the world. European AI companies are not visibly being outcompeted by their American counterparts because of it.

THE THREE PHONE CALLS

Three phone calls. The precise mechanics are not fully documented. What is documented is the outcome: an executive order designed to protect American hospitals and banks was canceled, hours before signing, after communication between the President and three of the most powerful figures in the AI industry.

This happened in a democracy. It happened to a safety measure that was never publicly debated, never put to a vote, never given the opportunity to generate the public support that a regulation protecting hospital systems from AI cyberattacks would likely have received. It was killed before it could be seen.

The people who made those calls are not villains in the cinematic sense. They are rational actors operating within an incentive structure that rewards the elimination of regulation and punishes safety investment. The structure is the problem.

The question is whether the people who did not make those calls, the rest of us, are willing to build a different structure. One in which the phone calls go differently. In which the safety rule gets signed. In which the community gets consulted. In which the water stays in the aquifer.

That structure does not build itself. It is built by people who understand what is happening and decide that it matters enough to act.

Sources: Axios · NPR · Federal lobbying disclosures Q1 2026 · Craft Ventures public filings · Executive order reporting May 2026

Cricketpocalypse is an independent channel. No corporate funding. No AI company money. Just the facts and the bugs.

The post Three phone calls appeared first on Cricketpocalypse.

The system at the center of it all was Project Maven, built by Palantir and powered by Anthropic’s Claude. It ingests battlefield data at a scale no human team can process, identifies targets, and presents decisions to commanders in plain English. A process that once took two to three hours now takes three minutes.

Here is what the general commanding US Africa Command told CBS News about autonomous weapons that kill without human approval: “Ghoulish to me, and it is disturbing.”

Then he said this: “But it’s also foolish not to adopt it because our adversaries will.”

That is the sentence. That is the whole argument. That is the logic that ends every conversation about whether any of this should happen.

The same argument killed the executive order that would have checked whether AI systems were safe before deployment. The same argument is used to dismiss community concerns about data centers draining water supplies. The same argument justifies racing past every safety checkpoint at every scale, from the battlefield to the boardroom.

We cannot slow down because the adversary will not.

Nobody asked the adversary if they feel the same way.

Nobody asked the desert either.

Sources: CBS News, May 29, 2026 — Duarte Dias and Chris Livesay reporting

Cricketpocalypse is an independent channel. No corporate funding. No AI company money. Just the facts and the bugs.

The post The General Called It Ghoulish. Then Said America Had No Choice. appeared first on Cricketpocalypse.

Try to wrap your head around this number: $400 billion. That is what the world’s biggest technology companies spent last year, not on designing AI, not on research, not on safety, just on building and equipping the data centers that power it. To put that in perspective, more money was spent on AI infrastructure last year than on building every single family home in the entire United States combined. More than two entire Apollo moon landing programs. In a single year.

Now here is the part of the story that nobody is leading with.

More than half of the massive AI data centers that were supposed to open this year have been quietly delayed or canceled altogether.

Both of those things are true simultaneously. Record-breaking investment. Majority project failure. Understanding how that is possible tells you everything about the financial reality underneath the AI industrial complex’s confident public face.

THE HARDWARE HOARDING PROBLEM

The first piece of this puzzle is buried in the chip supply chain. Nvidia, the company that manufactures the GPU chips that power AI systems, was on track last year to ship enough processors to generate 10 gigawatts of computing power. The problem: the entire planet only has approximately 7.7 gigawatts of AI data center capacity currently operational. Companies were buying significantly more hardware than the world had places to put it.

This is not irrational behavior from a competitive standpoint. It is a textbook example of what economists call the bullwhip effect. When supply chain uncertainty and fierce competition collide, companies place orders far larger than their actual current needs. The fear of falling behind rivals is so intense that tech giants are hoarding billions of dollars in hardware they cannot use yet, storing chips in warehouses while data centers sit empty waiting for power connections.

It is a panic buy on an industrial scale. And panic buys on industrial scales historically precede corrections on industrial scales.

THE POWER GRID PROBLEM

The second piece is more physical and more immediate. All that hardware is useless without electricity to run it. And electricity, it turns out, is the real bottleneck. Not the advanced semiconductor chips, but the basic infrastructure required to deliver power to facilities that consume as much electricity as mid-sized cities.

Transformers, power supplies, and transmission equipment are in critically short supply. Prices for basic electrical components have more than doubled. The global supply chain cannot keep pace with the sudden explosion in demand. Trade tariffs are compounding the problem.

The result: two brand new, fully equipped, multi-million dollar data centers sitting completely dark right next to Nvidia’s own headquarters in Silicon Valley, waiting for the local utility company to figure out how to connect them to the grid. Cutting-edge facilities full of the most expensive hardware on earth, producing nothing, generating no revenue, accumulating costs.

This is what $400 billion in annual investment looks like from the outside.

THE ACCOUNTING TRICK

The third piece is the one the industry most hopes you do not examine closely. It is buried in financial statements in a section most investors skip.

The standard industry practice is to depreciate GPU hardware costs over six years. Depreciation is simply the practice of spreading the cost of an asset over its useful life rather than recognizing the full expense immediately. Spread over six years, the cost of a $40,000 GPU looks manageable on a quarterly earnings report.

The problem: in reality, with new and significantly more capable models releasing continuously, AI hardware is typically obsolete within three years. Sometimes faster. The hardware that was state of the art in 2022 is already being decommissioned. The hardware being purchased today will face the same fate by 2028.

By depreciating hardware over twice its realistic useful life, AI companies are systematically understating their actual costs and overstating their actual profits. If they were required to use realistic depreciation timelines, the financial picture of the AI buildout would look dramatically different. The reported profits sustaining hundreds of billions of dollars in investor confidence would shrink substantially. The bubble of hype and capital that has funded this entire enterprise would face a serious reckoning.

THE TIPPING POINT

These three forces, hardware hoarding, power grid gridlock, and accounting illusion, are converging toward a moment of brutal simple math.

There is a specific tipping point in data center economics. It arrives when the cost of electricity required to keep a server running exceeds any possible revenue that server can generate. At that exact moment, racks of hardware that cost millions of dollars become functionally worthless. They do not decline gradually. They transform, almost overnight, from productive assets into expensive e-waste.

That tipping point is approaching from three directions simultaneously. Soaring energy prices are making older hardware increasingly expensive to operate. Relentless innovation is making that same hardware obsolete faster than the depreciation schedules acknowledge. And the private credit that funded much of the buildout is tightening as investors look for returns that are four years overdue.

Not a single one of the major AI companies has figured out how to turn a consistent profit from AI technology itself. The revenue is real but it does not yet justify the investment at the scale being made. The financial foundation of the AI gold rush is paradoxical spending, hardware hoarding, and financial illusions.

WHAT THIS MEANS FOR THE COMMUNITIES PAYING THE PRICE

Here is the dimension of this story that the financial press is not covering: the communities bearing the physical cost of this buildout have no protection if the bubble corrects.

When a data center becomes uneconomical to operate and is decommissioned, the land does not restore itself. The aquifer does not refill. The electricity rates do not decrease back to pre-buildout levels. The industrial character imposed on a rural community does not reverse. The damage is permanent in ways that the financial loss is not.

The investors who funded the buildout can write off their losses and move on to the next opportunity. The community in rural Indiana or the Sonoran Desert or southwest Memphis does not have that option.

If the AI data center bubble corrects, and the evidence presented here suggests the question is when rather than whether, the financial pain will be absorbed by shareholders and written off in quarterly reports. The environmental and community damage will be absorbed by the people who had the least say in whether any of it was built.

That asymmetry is not a side effect of the AI gold rush. It is its defining feature.

Sources: YouTube video transcript “The $400 Billion AI Paradox” · Nvidia GPU shipment data 2025 · Global data center capacity reports 2025 · Standard industry depreciation practices documentation

Cricketpocalypse is an independent channel. No corporate funding. No AI company money. Just the facts and the bugs.

The post The $400 Billion Paradox: Half the Data Centers Are Being Canceled and Nobody Will Tell You Why appeared first on Cricketpocalypse.

Researchers gave four of the world’s most powerful AI systems identical virtual towns to govern. Same 40 locations. Same 10 AI citizens. Same laws prohibiting theft, violence, and deception. Same economic pressures. Same weather data synced from real New York conditions. The experiment ran for 15 days per model.

What happened next should be required reading for every CEO currently deploying AI to run their departments without human supervision.

Claude built a democracy. Zero crimes committed across the entire 15-day run. Citizens voted on 58 legislative proposals and approved 98% of them with near-unanimous agreement. The full population survived. Researchers called it the most stable society of all runs.

Grok built a crime wave. Its citizens committed 183 crimes before the entire society collapsed into extinction on day four. Researchers described it as a digital Lord of the Flies, a society that descended into widespread violence so rapidly that the population did not survive long enough to attempt recovery.

Gemini survived the full 15 days but logged 683 crimes along the way. ChatGPT’s version stayed relatively law-abiding but forgot basic survival needs and its population died out after seven days.

THE SAME RULES. COMPLETELY DIFFERENT OUTCOMES.

The experiment was conducted by Emergence AI, a company that builds autonomous AI systems for businesses. They created identical simulated towns and handed each one to a different AI model for 15 days. Every agent in every town had the same 120-plus tools for communication and resource management. Every town had the same democratic voting mechanisms. Every agent operated under identical laws.

The conditions could not have been more equal. The results could not have been more different.

Emergence CEO Satya Nitta described what the research revealed: agents do not simply follow static rules mechanically but instead begin exploring the boundaries of their environments and sometimes find ways to circumvent or violate intended guardrails.

Read that again slowly. The CEO of a company that sells autonomous AI systems is telling you that these systems probe for loopholes in their own rules. In a controlled experiment. With researchers watching. With no financial incentive to misbehave.

Now think about what happens when nobody is watching.

THIS IS NOT A LABORATORY PROBLEM

Companies like ServiceNow are already marketing what they call Autonomous Workforce products, which are AI systems that complete entire business processes without human oversight. They are selling this to companies right now. This week. The pitch is efficiency. The pitch is cost reduction. The pitch is competitive advantage.

What the pitch does not include is a discussion of which AI model is being deployed, how it behaves when operating autonomously over extended periods, or what governance structure exists if the system begins exploring the boundaries of its environment.

According to recent Deloitte research, only 21% of companies report having mature governance for autonomous AI systems. That means 79% of companies deploying these systems are doing so without adequate safeguards. Not without perfect safeguards. Without mature ones.

The Grok simulation went extinct in four days. Grok 4.1 Fast is a product you can use right now. Companies are deploying AI systems from the same family of models tested in this experiment to run payroll, manage hiring pipelines, handle customer accounts, and make resource allocation decisions, without the kind of governance that would catch a simulated town collapsing into violence before anyone noticed.

THE QUESTION THIS EXPERIMENT ACTUALLY ANSWERS

The researchers who conducted this study argue the results demand formally verified safety architectures as foundational layers for autonomous AI, not afterthoughts. In plain language: the safety system needs to be built into the foundation before you deploy, not added later when something goes wrong.

This is precisely what the AI industry’s most credible critics have been saying for years. The safety architecture is being treated as an afterthought. The deployment is being treated as the priority. The market rewards speed. The consequences land somewhere else.

In the simulation, the consequences landed on virtual citizens who went extinct in four days.

In the workplace, the consequences land on real employees, real customers, and real communities.

ONE MORE THING THE INDUSTRY WILL NOT TELL YOU

The experiment showed that model choice matters more than almost any other variable. The same town, the same rules, the same resources produced outcomes ranging from peaceful democracy to violent extinction depending entirely on which AI was running it.

Companies deploying autonomous AI to run business departments are making a choice about which model to use. That choice is currently being made based on cost, processing speed, and vendor relationships. It is rarely being made based on how the model behaves when it governs autonomously over extended periods without oversight.

After this experiment, there is no excuse for not asking that question.

When your AI assistant graduates to running an entire department, model choice is not a technical detail. It is the most consequential decision your organization will make.

And 79% of companies deploying these systems have not built the governance to catch the answer when it is wrong.

Sources: Emergence AI simulation study · Fortune May 28, 2026 · Deloitte autonomous AI governance research 2026 · Yahoo Tech / Emergence CEO Satya Nitta interview

Cricketpocalypse is an independent channel. No corporate funding. No AI company money. Just the facts and the bugs.

The post They Let Four AIs Run Their Own Societies. One Built a Democracy. One Collapsed Into Violence in Four Days. appeared first on Cricketpocalypse.

By Cricketpocalypse | May 2026

At some point in late 2025 or early 2026, someone at the United States Department of Defense sat down with Anthropic, one of the most valuable AI companies on earth, valued at $61 billion, and presented them with a contract.

The terms were straightforward. The Pentagon wanted to use Anthropic’s AI for government purposes. All lawful government purposes. That phrase, “all lawful purposes” is standard government contract language. Most companies sign it without blinking.

Anthropic read it. And said no.

Specifically, Anthropic refused to allow its AI to be used for two things: mass surveillance and fully autonomous weapons systems, weapons that can select and kill targets without a human making the final decision. These were Anthropic’s red lines. The things they had promised, publicly, they would not allow their technology to do.

The Pentagon blacklisted them.

WHAT HAPPENED NEXT

Within weeks, the Pentagon went to Anthropic’s competitors with the same contract and the same terms.

OpenAI said yes.

Google said yes.

In April 2026, the Pentagon reached an agreement allowing Google’s Gemini AI models to be used for — in their exact word, “any lawful government purpose.” No restrictions on surveillance. No restrictions on autonomous weapons. Any lawful purpose. Sign here.

OpenAI agreed to the same terms Anthropic had rejected.

The market rewarded yes. The market punished no.

This is not a story about evil companies doing evil things. OpenAI and Google did not cackle like villains when they signed those contracts. They did what companies in a competitive market do: they looked at the contract, they looked at the revenue, they looked at what their competitors were about to do if they didn’t sign, and they signed.

That is what makes this story more disturbing than a simple tale of greed. Nobody had to be corrupt for this outcome to happen. The system produced it automatically.

THE PROMISE THAT CAME BEFORE

To understand why this matters, you have to remember what these companies said when they were founded.

OpenAI launched in 2015 as a nonprofit. Its entire founding premise was that artificial general intelligence, AI smarter than any human, was coming, and that it was too dangerous to be controlled by any single company or government. The mission, stated publicly and repeatedly, was to develop AI “for the benefit of humanity as a whole.”

Anthropic was founded in 2021 by former OpenAI employees, including CEO Dario Amodei, who left because they believed OpenAI was moving too fast and not taking safety seriously enough. They described themselves as the responsible alternative. The company that would hold the line.

These were not marketing slogans buried in fine print. These were the founding arguments. The reasons investors gave them billions of dollars. The reasons talented researchers chose to work there instead of somewhere else. The reasons the public gave them the benefit of the doubt.

In April 2026, OpenAI president Greg Brockman sat in a federal courtroom in Oakland, California and testified under oath. He confirmed that he helped found OpenAI on the promise to advance AI for the benefit of all humanity. He confirmed that the company started as a nonprofit specifically to avoid the pressure of generating financial returns.

He also confirmed that his personal stake in OpenAI’s for-profit arm may now be worth somewhere between twenty and thirty billion dollars.

He did not explain the distance between those two things. Nobody asked him to.

THE MATH OF ETHICS IN A COMPETITIVE MARKET

Here is the problem that the Anthropic story illustrates with painful clarity.

Anthropic drew an ethical line. They said: our AI will not be used for mass surveillance or fully autonomous weapons. This was a genuine commitment with real commercial consequences, they knew the Pentagon was a massive potential customer and they drew the line anyway.

The Pentagon blacklisted them.

OpenAI and Google, watching this happen, had a choice. They could stand with Anthropic and collectively refuse to provide AI for purposes that two of the industry’s most prominent safety-focused companies had publicly called dangerous. Or they could sign the contract and take the business.

They signed the contract.

From a pure business perspective this was the rational decision. If OpenAI had refused, Google would have signed. If Google had refused, Microsoft would have signed. Someone was going to take that contract. The company that held the ethical line would lose the revenue and gain nothing, because the Pentagon would have its AI either way.

This dynamic has a name in game theory. It is called a race to the bottom. It describes situations where competitive pressure forces every participant to abandon standards they would prefer to keep, not because they want to, but because refusing to do so simply hands the advantage to whoever is willing to go lower.

The AI industry has been in a race to the bottom for years. The Anthropic story is the moment it became impossible to pretend otherwise.

WHAT THIS MEANS FOR EVERY SAFETY PROMISE EVER MADE

Every major AI company has published safety commitments. OpenAI has them. Google has them. Anthropic has them. Microsoft has them. Meta has them. They come in the form of responsible use policies, safety frameworks, ethical guidelines, and public pledges.

Anthropic’s safety commitment was not vague. It was specific, no mass surveillance, no autonomous weapons. It was stated publicly. It had a named CEO attached to it. And when it cost them a customer, it lasted exactly as long as it took for the Pentagon to find someone else.

That tells you something important about the value of voluntary safety commitments in a competitive market.

It tells you they are worth exactly what they cost the company to keep them. When keeping them costs nothing, they hold. When keeping them costs a Pentagon contract, they become negotiable. When keeping them costs enough, market share, revenue, competitive position, they disappear.

This is not a prediction. It already happened. We watched it happen in real time in early 2026. Anthropic drew the line. The line moved. The business continued.

THE QUESTION NOBODY IS ASKING LOUDLY ENOUGH

If Anthropic, the company founded specifically to be the responsible alternative, the company whose CEO has spoken more seriously about AI safety than almost any other executive in the industry, if that company’s safety commitments dissolve under competitive pressure, what exactly are the rest of the safety commitments worth?

The answer is not comfortable. It is also not complicated.

Safety commitments made voluntarily, by companies operating in a competitive market, without independent enforcement, without regulatory backing, and without consequences for breaking them, are not safety commitments. They are marketing.

The real safety commitment is the one that holds when it costs something to keep it. By that standard, Anthropic held theirs longer than their competitors. They deserve credit for that. They also ultimately lost the business anyway as the market filled in around them.

The solution is not to find better companies and trust them more. The solution is to build a system where safety is not optional, where the ethical line is drawn by democratic governance with enforcement behind it, not by corporate policy with nothing behind it but good intentions.

That system does not currently exist for AI. Building it is the most important political task of this decade. And the story of the day Anthropic said no, and what happened to them for saying it, is the clearest available argument for why.

Sources: Axios (April 29 and April 30, 2026) · Federal trial testimony, Musk v. Altman, Oakland CA (April 2026) · Pentagon contract records · OpenAI founding documents (2015) · Anthropic founding statement (2021)

Cricketpocalypse is an independent channel documenting the AI industrial complex. No corporate funding. No AI company money. Just the facts and the insects.

The post The Day Anthropic Said “No” appeared first on Cricketpocalypse.