KI baute 16.000 Arbeitsplätze pro Monat ab und vergiftete die Nachbarn

Key Takeaways

• The backlash is bipartisan and accelerating: 57% of registered voters say Artificial Intelligence (AI) risks outweigh benefits, $64 billion in data center projects have been blocked or delayed, and 300+ state bills have been filed in just six weeks
• The externalized costs are staggering: AI data centers are projected to impose $20 billion per year in public health costs by 2028, the PJM grid auction added $9.3 billion in capacity costs onto ratepayers, and AI is cutting 16,000 net U.S. jobs per month
• The builders are the threat, not the opponents: The reckless speed of deployment, not anti-technology sentiment, is creating the conditions for a regulatory sledgehammer that could kill beneficial AI along with the harmful kind
• History is repeating: The current AI rollout mirrors the Gilded Age pattern of revolutionary technology deployed without regard for workers or communities, which triggered the Progressive Era’s regulatory backlash

57% of America Just Turned Against the AI Boom

Here is the math that should terrify every AI investor in Silicon Valley: 57% of registered voters now believe the risks of AI outweigh its benefits. Only 26% hold positive views. And 33% of voters say neither political party is equipped to handle it, meaning both parties are now incentivized to look tough on AI.

Meanwhile, $64 billion in U.S. data center projects have been blocked or delayed over the past two years by community opposition. That is not a protest movement. That is an industry-altering disruption driven by 142 activist groups across 24 states, a coalition that includes MAGA conservatives worried about property rights and progressive organizers fighting environmental injustice.

More than 300 data center bills have been filed across 30+ states in the first six weeks of 2026 alone, and Sen. Bernie Sanders has called for a federal moratorium on new construction.

The standard tech-industry explanation is that the public simply does not understand AI yet. The data tells a different story. The public understands AI’s costs perfectly well, because those costs are showing up in their power bills, their air quality, their job prospects, and their customer service experiences. The billionaire class is not losing the messaging war. They are losing the reality war.

Your Lungs: The $20 Billion Health Bill Nobody Voted For

AI data centers are not clean, quiet office buildings. They are industrial facilities running 24 hours a day, generating noise, heat, emissions, and water contamination. They are disproportionately sited in low-income communities and communities of color.

A February 2026 study published in Frontiers in Climate, the first peer-reviewed assessment of data center health impacts, found that AI-driven data center emissions are projected to cost the U.S. public health system more than $20 billion per year by 2028. The study examined Virginia’s Data Center Alley, the world’s largest concentrated hub, and documented harms across five vectors: air pollution from backup diesel generators (which pollute 200-600 times more than natural gas plants), excessive water use straining local supplies, chronic noise pollution, land use changes, and economic burdens on low-income households.

The damages are not hypothetical. In the Memphis area, Elon Musk’s xAI ran unpermitted gas turbines to power the Colossus supercomputer, bypassing the grid entirely. The company then installed 27 additional turbines across the state line in Southaven, Mississippi, for its Colossus 2 expansion. Residents in the nearby majority-Black neighborhood of Boxtown in South Memphis testified about a rotten-egg stench, worsening smog, and deteriorating respiratory health. A study by the Southern Environmental Law Center (SELC) found that xAI’s planned expansion to 40+ permanent turbines would impose $30-44 million in annual health damages on surrounding communities. The Environmental Protection Agency (EPA) updated its Clean Air Act standards for gas turbines in January 2026, which environmental groups allege puts xAI’s operations at odds with federal law, but only after the damage had already begun.

This pattern extends well beyond Memphis. Pollution from backup generators at Northern Virginia data centers drifts into Maryland, West Virginia, Pennsylvania, New York, New Jersey, Delaware, and Washington D.C. Cooling systems at some facilities use PFAS, or “forever chemicals,” linked to cancer, reproductive harm, and immune system damage. A single large data center can consume up to 5 million gallons of water per day, equivalent to the daily usage of a town of 10,000-50,000 people.

The communities bearing these costs did not vote for them. They did not benefit from them. And the companies imposing them received massive tax exemptions for the privilege. See the site’s prior coverage of the $1.6 billion AI tax dodge jacking up Virginia power bills and the Memphis Smokescreen for the full details.

Your Power Bill: $9.3 Billion in One Auction

Data center electricity demand is being subsidized by every residential ratepayer in 13 states, whether they use AI or not.

The numbers are precise. In the 2025-2026 PJM Interconnection capacity auction, which covers the power grid for 65 million people from New Jersey to Illinois, data centers drove 63% of the price increase, adding $9.3 billion in capacity costs that all ratepayers must absorb. The auction delivered an 833% jump in capacity prices from $28.92 to $269.92 per megawatt-day, the sharpest single-year increase in PJM’s 27-year history.

This translates directly to household budgets. Residential customers in the D.C. area saw bills rise by roughly $21 per month, with about $10 of that driven by the capacity market price spike. In Ohio and western Maryland, the increases ranged from $16-18 per month. In Virginia’s Data Center Alley, some residents experienced utility bill increases of up to 109% year-over-year.

The asymmetry is the scandal. Home electricity bills are skyrocketing while data centers negotiate preferential bulk rates. Virginia’s State Corporation Commission reduced Dominion Energy’s requested base-rate increases from $822 million to $565.7 million for 2026, and Oregon created a separate rate class to shield residential customers from data center costs. But these are defensive measures, states scrambling to contain damage that the industry created by building first and negotiating later.

Your Job: 16,000 a Month and Nobody Left to Buy the Product

The third externalized cost is economic, and it contains a paradox that should keep every AI executive up at night.

Goldman Sachs estimates that AI is erasing roughly 16,000 net U.S. jobs per month. The raw displacement is approximately 25,000 positions, partially offset by 9,000 jobs created through AI-augmented roles. Gen Z workers are absorbing a disproportionate share of the pain because they are concentrated in the exact administrative, customer service, and data-entry roles that AI automates most easily.

Goldman’s own economists caution that the net figure does not fully capture offsetting hiring in AI infrastructure: data center construction, power system upgrades, and chip manufacturing. Fair enough. But the structural problem remains. When a company can replace a $120,000-a-year mid-level analyst with a $20-per-month AI subscription, fiduciary duty makes the decision for them.

The paradox is this: who buys the product if nobody has a job?

AI companies are building tools priced at $60-200 per month for heavy professional use, based on publicly listed pricing from Cursor, Claude Code, and GitHub Copilot tiers. These subscriptions depend on a workforce of knowledge workers who can afford them and whose employers fund them. Every job AI eliminates is a potential customer destroyed. Every customer destroyed is revenue that will never fund the next model. The “democratization of technology” that AI companies promised is arriving at a price point that excludes the very people being democratized out of employment.

The irony compounds. An estimated 85% of AI startups are expected to fail within three years, according to industry advisors tracking the shakeout. The survivors, likely the same handful of foundation-model companies, will consolidate into an oligopoly. An oligopoly charges more, not less. The technology gets less accessible over time, not more.

This is the observable trajectory: concentrate the gains, externalize the costs, and hope the market holds up long enough to reach profitability. It is a bet that the economy can absorb the shock. History suggests otherwise.

Your Trust: 600+ Court Cases and an Accountability Vacuum

The fourth externalized cost is the one the industry discusses least: AI does not work reliably enough to replace the humans it is replacing.

There are now more than 600 court cases globally involving AI-generated hallucinations, outputs that are confidently stated, completely fabricated, and presented as fact. In the legal profession alone, 128 lawyers have been implicated for filing AI-hallucinated content. In February 2026, there were 33 judicial opinions involving lawyers caught submitting fabricated AI citations, more than one per business day.

For consumers, the consequences are immediate and personal. Nearly one in five people who have used AI-powered customer service report zero benefit from the experience. That failure rate is almost four times higher than for AI use in general. Chatbots hallucinate refund policies, fabricate product specifications, and make promises that no human representative would authorize.

The legal precedent is already set. In Moffatt v. Air Canada, a tribunal ruled that the airline was liable after its AI chatbot hallucinated a bereavement refund policy that did not exist. The court’s logic was blunt: a company is responsible for all information on its website, whether generated by a human or a bot. If an AI promises a refund, the company owes the refund.

But most companies replacing human customer service representatives with AI chatbots have not built the accountability infrastructure for when those chatbots fail. They have no audit trail for hallucinated advice. They have no escalation path for customers harmed by fabricated information. They have externalized the quality-control function onto users who do not know they are talking to an AI, cannot evaluate whether the output is accurate, and have no recourse when it is not.

The Federal Trade Commission’s (FTC) “Operation AI Comply” signals enforcement intent, but the regulatory infrastructure lags years behind deployment. Businesses are cutting headcount, pocketing the savings, and leaving customers to absorb the cost of unreliable AI in time wasted, money lost, and trust destroyed.

The Gilded Age Is Back

Every cost described above follows a pattern that American history has seen before.

In the original Gilded Age (1870s-1900s), railroad barons and steel magnates built transformative infrastructure. The railroads connected the continent. Steel built the skyscrapers. Oil lit the cities. These were genuine technological revolutions that reshaped civilization.

They also killed workers in unsafe mines, poisoned rivers with industrial waste, manipulated commodity prices, and bought legislatures. The richest 0.0001% now control a greater share of wealth than they did in the original Gilded Age. The industrialists argued that the public simply did not understand the benefits of progress, that the costs were temporary and the gains permanent.

The public responded with the Progressive Era: antitrust law, the FDA, labor protections, the income tax. Not because Americans opposed technology, but because unchecked deployment imposed intolerable costs.

The 2026 parallel is unmistakable. AI barons are building genuine technological marvels while externalizing every cost onto communities, ratepayers, and workers. The bipartisan moratorium movement, 55% Republican and 45% Democrat according to Data Center Watch, is the modern equivalent of the Progressive coalition. The question is not whether regulation is coming. The question is whether it arrives as a scalpel or a sledgehammer.

The industry’s behavior is determining the answer. Every unpermitted turbine in Memphis, every $9.3 billion rate increase dumped on ratepayers, every hallucinating chatbot left unsupervised, every 16,000 jobs erased without a transition plan pushes the political system toward the sledgehammer.

The Path Not Taken

None of this is inevitable. The technology itself is not the problem. The deployment model is.

Research from MIT Sloan shows that AI augmentation, using AI to enhance human productivity rather than replace workers, delivered a 14% average productivity gain in a study of customer support agents at a Fortune 500 firm, with improvements of up to 35% for the least experienced workers. Broader analysis suggests only about 12% of current jobs face full task automation. The remaining 88% are candidates for augmentation that makes workers more productive, not obsolete.

The healthcare applications are real. AI pathology models achieve approximately 94% accuracy across 19 common cancer types, according to a 2025 study in China. AI documentation tools cut medical record-keeping from 10-15 minutes per case to roughly one minute. Multilingual conversational AI provides frictionless care access for patients who do not speak English.

Data centers can be sited responsibly. Noise barriers work. Closed-loop water systems exist. On-site renewable generation eliminates diesel backup. Separate rate classes protect residential ratepayers. Community benefit agreements ensure local economic value. Oregon, Virginia, and Pennsylvania have already begun implementing these solutions.

But responsible deployment is slower and costs more upfront. It requires negotiation with communities rather than bulldozing through them. In a race where every quarter of delay means a competitor ships first, the incentive structure punishes responsibility.

That is the structural flaw. Not evil. Not conspiracy. Just a market dynamic where the costs of speed are borne by everyone except the people making the decision to go fast. The tragedy is that the recklessness is self-defeating: the $64 billion in blocked projects, the 300+ state bills, and the 57% disapproval are the market’s antibodies. The industry is generating its own immune response, and that response is getting stronger every quarter.

AI could be the most beneficial technology of the 21st century. It could augment workers instead of replacing them, power itself cleanly, compensate the communities it inhabits, and build trust through reliable performance. That version of AI would face no backlash, because it would not deserve one.

The version being built in 2026 deserves exactly the backlash it is getting.