John Sherman traveled to Holly Ridge, Louisiana - a town of 2,000 people - where Meta is building the world's largest data center. No resident was informed. Here is what he found on the ground.
Every time you run an AI query, generate an image, or open a chatbot, that request travels to a physical building. A very large physical building, filled with servers, drawing enormous amounts of power and water, built somewhere specific - on someone's land, next to someone's home, in someone's town.
Most people never think about where.
John Sherman, host of For Humanity and president of GuardRailNow, decided to go find out. He flew to New Orleans and drove four and a half hours north into rural Louisiana to a place called Holly Ridge - population 2,000 - where Meta is building Hyperion, described as the world's largest data center. Larger than Manhattan.
He knocked on doors. He sat in living rooms. He looked at the water.
What he found should concern anyone paying attention to where AI development is actually heading.
Meta's Hyperion project is one of the largest single AI infrastructure investments in history. The facility is being constructed in Richland Parish, Louisiana, in and around the small rural community of Holly Ridge.
The site was selected without public announcement. Louisiana passed legislation explicitly allowing the state to withhold information about incoming development projects from local communities. Town officials in Holly Ridge were reportedly bound by NDAs before construction began - legally prohibited from informing their own constituents.
The result: the first thing most Holly Ridge residents knew about Hyperion was when the construction trucks started arriving one day. No letter. No town hall. No vote. No warning of any kind.
This is not an isolated case. Hundreds of communities across dozens of states are currently organized against data center projects following the same pattern: rural or low-income area, secrecy until construction begins, then a press release about jobs and economic growth. In Indianapolis, residents found out in time, organized, and blocked a major Google data center. The difference between Indianapolis and Holly Ridge is not politics. It is access to information - and that information gap is by design.
John Sherman's visit to Holly Ridge produced firsthand testimony that goes well beyond policy abstractions.
Residents across multiple households described tap water that runs brown like coffee some mornings and smells like bleach on others. Several now buy bottled water by the case - not by preference but by necessity. One elderly woman, who has lived on her land since 1969, gives her dog spring water because the tap water made the animal sick. She now uses an inhaler she never needed before. Her family has stopped visiting.
Construction dust is constant. John's team could feel it in their noses and throats after just a few hours on site. Residents have been living with it for over a year.
Thousands of fully loaded 18-wheelers travel through Holly Ridge every day, including past the local elementary school. Roads have been torn up. Driveways have been driven over without permission. A neighbor's animal was killed by a construction truck.
Not one resident was consulted before any of this began.
Amazon previously settled for $20 million in Morrow, Oregon over nitrate contamination linked to a data center. The pattern of harm - water quality degradation, dust, traffic, noise, health impacts - is documented and recurring. It is not unique to Louisiana.
For years, the technology industry described its infrastructure using language designed to suggest weightlessness. The cloud. Seamless. Invisible. Everywhere and nowhere.
The Hyperion data center in Holly Ridge is not weightless. It is concrete and steel covering an area larger than Manhattan. It draws enormous volumes of water and electricity. It generates construction dust that coats the lungs of elderly residents. It sends thousands of trucks past a school every day.
Every AI product from Meta - every Instagram algorithm, every recommendation engine, every generative AI feature - runs on infrastructure like this. Built in places like this. Paid for, in part, by the communities that host it: through degraded water, damaged roads, health impacts, and the permanent alteration of places people chose to live.
Understanding AI risk means understanding this. The technology is not abstract. Its costs are physical, local, and concentrated - almost always in communities with the least political power to resist.
During the drive to Holly Ridge, John Sherman debated this question directly with the team from Tail End Films, who are producing a documentary on the AI race called Making God.
The counterargument is reasonable: the AI safety movement is small. Resources are limited. Communities experiencing data center harms may not be the same audiences most receptive to arguments about long-term extinction risk. Is this the right fight for a movement with bigger existential concerns?
John's answer: yes. For two reasons.
First, data centers are where ordinary Americans first encounter - and first resist - the AI future being built around them. One resident in Holly Ridge, unprompted and with no background in AI safety, told John: "I think it's going to eventually get too smart for its own good, and it's going to take us over. And then what are we living for?" She arrived at that concern herself. Because she is living next to the machine.
Second, the same pattern that produced Holly Ridge - concentrated power, absence of consent, externalized costs, regulatory capture - is the same pattern that makes AI governance so difficult at every level. Fighting it locally is not a distraction from the larger problem. It is practice for the same fight at a larger scale.
Several Holly Ridge residents invoked Louisiana's long history with extractive industry without any prompting. Oil came in, money flowed to people elsewhere, the infrastructure aged, the companies moved on, and the communities were left with the aftermath.
They predict the same arc for data centers.
"In 15 or 20 years, it'll sit empty," one resident said. "Just like an old Walmart. They'll move on."
This is not cynicism. It is pattern recognition from people who have watched it happen before. The energy demands of today's AI hardware are so intensive that the current generation of data centers may become obsolete within a decade as chip architectures shift. The communities that hosted them will be left with damaged roads, altered water tables, higher electricity costs, and facilities that no longer serve any purpose.
Some projections suggest electricity costs in regions with high data center density could double as facilities bid against residential users for power. The communities that bear these costs are rarely the ones who capture the economic benefits.
If this matters to you, there are concrete things worth doing.
Advocate for federal standards requiring community notification and consent processes before data center construction begins. Support legislation that mandates environmental impact assessments for AI infrastructure projects. Follow the reporting of journalists like Drew Hawkins at the Gulf States Newsroom, who are documenting these impacts at the local level.
And understand what AI development actually requires - not as an abstraction, but as a physical reality with physical consequences for real people in real places.
Watch the full episode: For Humanity #85 on YouTube
Take action on AI risk: https://safe.ai/act
Support GuardRailNow's work making AI risk visible: guardrailnow.org/donate
