Then the prediction, with odds and grading dates. And at the end, a caveat for investors about what a wave of bad sentiment is actually worth.* **By Nicholas Thomas · Sunday, September 6, 2026 — TrendyVest Editorial. Opinion, with receipts, fact-checked before publication.
Filed 2:45 PM ET. Not investment advice.** --- There are two stories running through America's data center boom right now, and if you treat either one as the whole story you get the issue wrong. The first is uncomfortable, and it's documented.
Foreign-linked influence operations have identified American opposition to AI infrastructure as a useful political fracture, and they've tried to widen it. The threat-intelligence firm Alethea, after six months of monitoring, wrote that the topic "has been elevated within Russia's playbook into a recognized domestic fracture point." Read that once more. Not a policy outcome.
A fracture. The goal isn't to win the argument about electricity bills in Box Elder County, Utah. The goal is for the argument never to end, for the permits never to issue, and for the country that invented the technology to spend the decisive decade of it arguing with itself about cooling towers.
The second story matters every bit as much. Americans have concerns of their own, and they're legitimate. Electricity costs, water, noise, transmission lines, secrecy, land use, and who pays for the wires are real policy questions.
A Chinese bot didn't invent them. Seven in ten Americans oppose an AI data center in their area, per Gallup, and that number didn't come from Beijing. Both of those things are true at once, and that's exactly why this is a national security problem rather than a zoning dispute.
The fracture is happening. Not because the bots are winning — I'll show you the evidence that they mostly aren't — but because we're doing the job for them, county by county, on a set of beliefs about data centers that the receipts from the places that actually have them don't support. Fourteen statehouses proposed moratoriums this year on the one piece of infrastructure the AI race is actually decided by.
The United States doesn't have to choose between letting hyperscalers build anything they want and refusing to let data centers be built at all. There's a third option, and it isn't complicated. Make developers pay their way.
Protect the watersheds that can't spare the water. Require disclosure. Set noise and siting standards.
Demand a credible power plan. Capture the tax base. Build the grid.
And keep building the compute this country needs — because while Americans argue about the rules for building AI infrastructure, China is building enormous quantities of energy infrastructure with no argument at all. That's the real race. I'll make the prediction with odds, tell you what would make me wrong, and, at the end, tell you what I think the fear is worth. **THE CAMPAIGN, ITEMIZED.** Three investigations this year, three different methods, one conclusion.
Read each precisely, because the imprecise versions are already circulating. *OpenAI, June.* The company banned a cluster of ChatGPT accounts it assessed as likely originating in China and named the operation "Data Center Bandwagon." The operators had prompted the model in Simplified Chinese "while repeatedly asking for English- and Chinese-language outputs, posing as Americans from a variety of backgrounds," and the output went onto X, Facebook and YouTube. Its central narrative was that data centers and AI were "increasing electricity demand and causing higher costs for ordinary Americans." OpenAI's attribution went a step further than most: the operators were "likely part of a social media operations team at a private Chinese technology company conducting work for Chinese provincial-level government clients." A second cluster criticized U.S. tariffs and, for good measure, spread a false claim that ChatGPT user data had been breached. That isn't speculation about foreign influence.
It's a platform disruption, written up by the platform. But here is the sentence that matters just as much, and it should stay in bold type: OpenAI **"found no evidence of meaningful breakout beyond its own activity,"** and said the operation "does not appear to have shifted public opinion." Foreign actors tried to exploit the debate. There is no evidence they created the American backlash. *Alethea, July 9.* Six months, January through June, three governments. **China:** "roughly 46 Chinese state media articles pushed a triumphalist counter-narrative," including a CGTN segment on American electricity bills that was itself AI-generated, a China Daily piece headlined "AI boom sends electricity bills in US skyrocketing," and satellite-imagery explainers under the headline "AI infrastructure is moving to residents' doorsteps" — one of them, per the New York Times' account of the same research, run by a Chinese state newspaper about a data center in Gainesville, Virginia. **Russia:** RT aired a segment titled "Is resistance futile: Local communities take the fight to high-tech data centers"; Portal Kombat, "a Russian-linked network of hundreds of sites that publish in dozens of languages," laundered a "Data Centers against Humanity" message; the Storm-1516 operation fabricated two stories about a U.S.-linked data center in Hrazdan, Armenia, one alleging the town "faces a magnitude-7.4 earthquake within 35 days," the other falsely attributed to a real tech journalist; the pro-Kremlin channel Rybar dismissed OpenAI's disclosure as "organic" resistance; and fake domains dressed as American tech press — *tech-crunch.org*, *gizmodo.cc* — carried the goods. **Iran:** Mehr, Fars and Press TV tied data centers to pollution and to Israel.
Underneath the governments sits a layer Alethea calls AI slop: Facebook pages, "many of which are anonymously-run and based in Bangladesh yet carry seeming innocuous names such as 'Life in Oklahoma' and 'Only in Montana,'" sharing AI-generated imagery, and a TikTok cluster of machine-voiced "breaking news" videos, "a hundred of them, served algorithmically to the same zip code." The one message every actor converged on, in Alethea's words: *"America is hoarding power and water for the elites while ordinary people suffer."* *X, August 27.* On Thursday night the platform's safety team said it had investigated a suspected Chinese influence network of roughly 200,000 accounts, within which about 200 were "posting in a manner that could manipulate a legitimate debate about American AI and energy policy." The content, per Axios: "how AI strains the electricity grid and increases utility prices," plus "cartoons that depicted data-center operators enriching themselves at the public's expense." Precision matters here, because the number is already being misquoted. It is not accurate to say 200,000 accounts were campaigning against American data centers. The finding was a 200,000-account suspected network, inside which X identified about 200 accounts working this debate.
That's still significant. It's just a different, and more defensible, claim. Former Rep.
Chris Stewart, founder of Public First Action, to the Deseret News: *"It's not just 200; it's probably 200,000. China made the decision that they could win this race if they could convince the American people to oppose data centers, which is the compute power we need."* Sen. Dave McCormick of Pennsylvania, on X: *"They want us to lose.
They want America divided, afraid to build, and dependent."* Stewart may be right about the intent. The evidence supports the smaller number, and I'd rather be right about the evidence. Now the honest part, which the people who want this to be a simple story won't give you.
Jim Prosser, a Silicon Valley communications consultant who once ran Twitter's corporate communications, told Axios: *"If you can get both Greg Abbott and Kathy Hochul agreeing on something, it's probably not a Chinese psyop."* Taylor Barkley, the Abundance Institute's director of federal government affairs: *"I don't think that this release should in any way undermine the real concerns Americans have about data center buildout."* Alethea itself wrote that *"the line between foreign propaganda and domestic political messaging is increasingly dissolving at the level of form and technique"* — the fakes and the real thing now look alike, and the real thing is enormous. Gallup, March: seven in ten oppose an AI data center in their area. A University of Pennsylvania survey in early August, per Axios: 61 percent oppose one locally.
The Economist/YouGov: 24 percent say new data center construction benefits the country, 47 percent say it's detrimental. The partisan split, per the Christian Science Monitor: 75 percent of Democrats and 63 percent of Republicans oppose one in their community. This is the most bipartisan thing in America.
No bot farm did that. So the strongest conclusion the evidence supports is this: actors linked to China, Russia and other adversarial information ecosystems have recognized America's data center controversy as an exploitable division and have tried to amplify it, and the available evidence does not establish that they created the underlying opposition or moved public opinion. That's enough.
We don't need to embellish it, and the people embellishing it are doing the adversary a favor, because an exaggerated claim is an easy one to dismiss. And here's how I'd have you read the campaign, because it's the reading that makes this a security story rather than a spy story. Influence operations don't invent grievances.
They audit the ones you already have and pour accelerant on the ones that will burn. Look at which claims every foreign actor picked: electricity bills and water. Not noise, which is real.
Not the secrecy of the deals, which is real. Not housing, which in Abilene, Texas is a crisis. They picked the two claims with the *weakest* evidence behind them, because those are the ones that need help.
That's the tell. It's why the middle of this editorial is a fact-check rather than a sermon, and it's why the fix for the influence campaign and the fix for the grievances are the same fix. **THE REVOLT, BY THE NUMBERS.** Data Center Watch counted **at least 75 projects worth about $130 billion blocked or delayed in the first quarter of 2026** — "roughly matching the scale of all of 2025 in just three months," the largest single-quarter total on record; in Virginia alone, an estimate cited by Axios puts delayed projects at $45.8 billion. Active opposition groups went from 396 at the end of 2025 to **833 by March**, in 49 states.
More than 300 state data center bills were filed in the first six weeks of the year. Fourteen states saw statewide moratorium proposals, from both parties. Note the verb: fourteen states did not enact moratoriums.
The movement got big enough that proposals for statewide pauses appeared in fourteen statehouses, which is a different thing. Then the governors moved. Maine's Legislature passed the first statewide moratorium in April, on facilities of 20 megawatts or more, through November 1, 2027.
Gov. Janet Mills vetoed it on April 24, and on April 29 the House fell short of the two-thirds needed to override, 72 to 65 — I'll come back to her reasons, because they're the case. New York's Gov.
Kathy Hochul signed Executive Order 62 on July 14: a one-year halt on state environmental permits for hyperscale data centers while the state writes standards and what it calls a Community Investment Framework, citing development that threatens to "hike up utility bills, deplete our natural resources, and create uncertainty." On August 3, Texas's Gov. Greg Abbott paused approvals for new data centers in ERCOT's interconnection queue pending an audit, and ERCOT stopped its "batch zero" review. Pennsylvania's Gov.
Josh Shapiro signed an order on August 18 requiring local consent, a "bring your own power" standard, local hiring and a binding consent agreement. Ohio's ballot amendment to ban data centers above 25 megawatts collected about 70,000 of the 413,000 signatures it needed and is aiming for 2027; Port Washington, Wisconsin voted 66 percent to require voter approval for any data center tax incentive; Frederick County, Maryland verified a referendum to overturn a 2,600-acre approval. Some of that is prudent, and I'll say so plainly.
Texas is the case in point. Requests to ERCOT went from about 48 gigawatts in 2023 to **474 gigawatts** in 2026 — more than 1,800 projects, about 90 percent of them data centers, against an all-time peak demand of 85,508 megawatts — and only 28 of 377 notified companies had bothered to answer the state's survey. Reuters reported this week that the Midwest, Mid-Atlantic and South together hold more than 700 gigawatts of requests, more than ten times what every data center in America consumes today, and that regulators now have a word for the duplicated, speculative and unfinanced part of it: ghost demand.
Texas PUC Chairman Thomas Gleeson: *"When you don't know what is real, you really don't know how to build the infrastructure for it."* A queue like that is speculation, not demand, and a regulator that pauses to sort it is doing its job. America shouldn't build transmission lines for phantom projects. I'll say the same for consent agreements and disclosure.
But there is an enormous difference between *prove your project is real, pay for the infrastructure it requires and meet community standards* and *don't build the infrastructure at all.* The first strengthens the buildout. The second is the fracture point, adopted as policy. **THE RECEIPTS FROM THE TOWNS THAT SAID YES.** This is the part of the story the campaign needs you not to read, so read it slowly. *Loudoun County, Virginia* is the most important case, because it has lived with data centers at a scale nobody else has. About 53 million square feet of them — 920 football fields — paid **$1.1 billion in local taxes** in the year referenced by the county's FY2026 budget, **38 percent of the General Fund.** A decade ago the computer-equipment tax brought in about $150 million.
Over that decade the county cut its real property tax rate from **$1.145 per $100 of assessed value in 2016 to $0.805 in 2026.** That doesn't mean every homeowner's bill fell — assessments move too — but the rate fell by nearly a third while data centers became the county's fiscal base, and the money built the Ashburn Recreation and Community Center, 50-meter pool included, which opened in July 2025, and funds schools, roads and emergency services. Northern Virginia's data centers as a whole paid about **$1.3 billion in property taxes in 2024**, per the Northern Virginia Technology Council's report by Mangum Economics, which also estimates a $40 billion statewide economic impact, more than 112,000 jobs supported and $1.5 billion in annual state tax revenue. That report is industry-commissioned, so weight it accordingly.
Its counterfactual is the number I'd remember anyway: without data centers, Loudoun's rate would have to be **$1.537** instead of $0.805 to raise the same revenue, which on an $800,000 home is the difference between a $6,440 bill and a $12,296 one — about **$5,856 a year**, by the Chamber of Progress's arithmetic on the NVTC figures. Prince William's rate would need to rise 29 percent; Culpeper's, 34. Now the independent number, because I don't want you taking the industry's word for it.
Virginia's legislative watchdog, JLARC, studied the sector in December 2024 and found **74,000 jobs, $5.5 billion in labor income and $9.1 billion in annual GDP**, with data center revenue running as high as 31 percent of total local revenue in mature markets, and found that data centers currently pay their allocated share of costs under existing rate structures. Then it added the qualification that matters: "most of these economic benefits derive from the construction phase rather than data centers' ongoing operations." A typical 250,000-square-foot facility "may have approximately 50 full-time workers, about half of which are contract workers," while "at the height of construction, approximately 1,500 workers are on site." It also found the cost side, and so will I: the state's sales tax exemption was worth $928 million in FY23; a typical Dominion residential customer could see generation and transmission costs rise $14 to $37 a month by 2040 if nothing changes; a third of facilities sit near homes; and low-frequency noise rarely violates an ordinance but does wear on the people who live with it. That's what a real ledger looks like.
Both columns. *DeKalb, Illinois.* Meta's campus has generated **$72 million in additional property tax revenue since 2022.** About 60 percent goes to the schools — just under $32 million in the first three years — and the district opened the **$34 million Dr. Leroy A. Mitchell Elementary School** in August 2025 for 400 students.
Residents' share of the school budget fell from more than half to about 30 percent; the city's assessed valuation went from roughly $600 million to close to $2 billion; homeowners' taxes were held flat while their property values rose. City Manager Bill Nicklas, to St. Louis Public Radio: *"The economic impact has been beyond our wildest dreams."* Forty miles away in Aurora, a city that taxes data centers by the megawatt — about $1.078 million a year per facility — spends the money cautiously, writes decommissioning requirements into its ordinances and has fielded noise complaints about rooftop fans.
Two towns, one lesson: the money is real, and so are the fans. And none of it proves every agreement will turn out this way; tax structures, incentives and power arrangements all differ. But it destroys the idea that communities necessarily receive nothing. *Jay, Maine.* The governor who vetoed her own state's moratorium explained why in writing, and she started by conceding the other side's point: a moratorium "is appropriate given the impacts of massive data centers in other states on the environment and on electricity rates." She vetoed anyway, because the bill didn't exempt a project on the site of the Androscoggin Mill, whose 2023 closure "dealt a devastating blow to the Town of Jay and its surrounding area." The project would "create more than 800 construction jobs, at least 100 high-paying permanent jobs, and would contribute substantial property tax revenue to the Town of Jay." *"I supported the exemption and would have signed this bill if it had included it."* A Democratic governor, in a state whose Legislature had just passed a ban, explaining that the ban would take a paper-mill town's second chance away.
That is the entire national debate compressed into one town: protect communities, and understand what saying no costs them. *Abilene, Texas* shows the other side of the ledger. Stargate — OpenAI and Oracle's flagship — has meant **9,000 construction jobs and, per Oracle, 1,000 permanent ones**, on an 85 percent tax abatement that still yields the city an estimated **$30 million a year.** Company figures, so label them. And here is the cost, at full size, because a column that hides it isn't worth your time: average rent in Abilene is about $2,600 a month, **up nearly 50 percent in a year**; it's cheaper to rent in Austin; a shelter case manager told KUT he hadn't had children in his shelter in a long time, until now; Abilene Christian University had more than 100 incoming students without housing.
That's a boomtown, and boomtowns break things. It's an argument for competent local government and housing policy. It isn't evidence that the investment has no value. *Box Elder County, Utah* is the best example of why neither side should get a blank check.
The Stratos Project — Kevin O'Leary's O'Leary Digital, 20,000 acres, down from 40,000, comparable in size to Manhattan, 9 gigawatts at full build, near the Great Salt Lake — is the one where the state's own FAQ and the opposition's experts describe two different planets. The state, verbatim: "The development will produce all power on site; it is stand-alone power that will not add pressure to the grid." It "would use a closed-loop chilling system combined with dry (air-based) cooling." "The systems use only existing water rights attached to private property, which means the project will have lower net consumption than current agricultural or ranching use." "All public infrastructure for the project area will be paid for by the developer, not county taxpayers." The county "is expected to receive $30 million in new revenues annually" in the initial phases, rising to as much as $108 million at full build-out, with a developer commitment to "a projected 2,000 permanent jobs." The critics, per the Monitor: Robert Davies, an associate professor of physics at Utah State, calculates the thermal output could equal "23 atomic bombs" of waste heat a day and raise nighttime temperatures by 12 degrees; Ben Abbott, a BYU ecologist who runs Grow the Flow, estimates the water need could reach 100,000 acre-feet a year. Both sides are, today, unproven.
The water-rights application filed March 25 was withdrawn May 6; the state hasn't published the engineering behind "lower net consumption"; I haven't seen the model behind the 12 degrees. A June poll had the project at 26 percent support and 60 opposed, which tells you who wins an argument when nobody shows the numbers. Gov.
Spencer Cox: *"We can't just say no and shut the doors... let China win this, this technology race."* He has since conceded that "the process was not good" — the Military Installation Development Authority approved the project in May without the usual community input, and the county denied a referendum. And one line from the state's FAQ that hasn't made it into the fight: "The Undersecretary of the Air Force asked MIDA to find locations for independent energy and computing power because supporting energy resilience, computing power and data storage is critical for defense operations." The Pentagon asked for the site. The process that answered it was bad.
Both are true, and only one argues for a ban. The correct response isn't yes or no. It's *show your work* — the power plan, the water rights, the cooling design, the heat model, the infrastructure costs, the tax agreement, the decommissioning plan.
Then decide. **THE EIGHT CLAIMS, GRADED.** Plain language, sources named. Where the opponents are right, I say so. *1. "Data centers are raising your electric bill."* **Too simple as a national claim; true in one specific market, with the fix already written.** Lawrence Berkeley National Laboratory's October 2025 analysis, updated this spring, found the main drivers of rising rates are poles-and-wires investment and extreme-weather costs, and that load growth correlates with *lower* rates, because fixed costs get spread over more kilowatt-hours.
EPRI's 2026 study found data center growth associated with **6 percent lower average retail rates from 2019 to 2024**, with the caveats that the period ends before the current surge and that if the AI demand doesn't show up, the remaining ratepayers eat the fixed costs. Columbia's Center on Global Energy Policy — Google-supported, so weight it — found rates jumped 6 percent in 2025, twice inflation, blamed planning, utility incentives, permitting and supply chains, and found that the places with the highest demand growth saw inflation-adjusted rate *declines*, with the caveat that the outcome isn't automatic. Its conclusion isn't "data centers lower everyone's bill." It's that load growth alone doesn't decide the outcome; how new generation and transmission get planned and paid for decides it.
The exception is real: PJM, the mid-Atlantic grid, where the independent market monitor says data centers accounted for **$6.3 billion of the $16.4 billion** in the latest capacity auction — 38 percent — and 46 percent of the last four auctions' $63.6 billion combined. Joe Bowring, the monitor: *"You have to open your eyes and recognize that it is really a paradigm shift, and failing to do that imposes costs on other customers."* His fix is a separate auction for data center load on 15-year contracts. That is the right fight — cost allocation — and the industry has already conceded it on paper.
The Ratepayer Protection Pledge, signed at the White Ho