Five hundred people showed up to a county fairgrounds in northern Utah last week to stop something. They’d filed 2,300 formal protest filings with state engineers. They came with signs. When the three commissioners voted yes anyway — unanimously — the crowd shouted “Shame! Shame! Shame!” as the officials left the room.
The commissioners walked out of their own meeting.
What they’d approved is called the Stratos Project: a data center campus on 40,000 acres of desert land in Box Elder County, backed by Shark Tank investor Kevin O’Leary. At full buildout, it would run on 9 gigawatts of power — nearly what New York City uses — generated entirely from natural gas, on-site. Utah currently uses about 4 gigawatts total. The Stratos Project would more than double the state’s total electricity consumption and increase its carbon emissions by roughly 50%.
Bryce Canyon National Park is smaller than what they just handed over.
I want to talk about who’s responsible for this. Because there’s a version of that conversation that lets the people actually building this stuff off the hook — and it’s been running for decades.
Edward Abbey’s Beer Can
Edward Abbey, the desert anarchist and author of Desert Solitaire, once wrote: “Of course I litter the public highway. Every chance I get. After all, it’s not the beer cans that are ugly; it’s the highway that is ugly.”
Abbey wasn’t arguing that littering is fine. He was making a harder point: the beer can is a distraction. The real ugliness — the real environmental violence — is the highway carved through the wilderness in the first place. Focus on the can, and you’ll spend your life picking up after a system that will keep producing cans faster than you can collect them.
This argument has a long history in environmental politics. In the 1950s, Coca-Cola, Anheuser-Busch, and the packaging industry quietly funded the “Keep America Beautiful” campaign — the one with the crying Native American watching someone throw trash from a car window. The campaign was brilliant. It shifted the entire conversation about pollution from industrial production to individual littering. It worked. It’s still working.
The AI version of this playbook is already running.
The Highway Being Built Right Now
The Stratos Project isn’t just a private investment in a county fairgrounds. It was routed through MIDA — Utah’s Military Installation Development Authority — a state entity specifically designed to fast-track projects with national security implications, bypassing many of the local approval processes that would normally apply. National security. That’s the framing O’Leary’s team chose.
That framing matters. Because once a project becomes a matter of national security, community objection starts to look like obstruction. Five hundred people shouting in a fairgrounds becomes noise. Two thousand three hundred water filings become bureaucratic delay. The highway gets built.
And Box Elder is not alone. This is the pattern.
The AI industry — led by Microsoft, Google, Amazon, Meta, and a growing field of sovereign wealth funds and defense contractors — is in the middle of the largest infrastructure build-out in modern history. Goldman Sachs has projected that data center power demand will triple by 2030. The US Department of Defense has been among the biggest AI investors for years. Countries competing to lead in AI are treating data center construction the same way previous generations treated the space race or nuclear infrastructure: as a strategic imperative that overrides local concerns.
In Indonesia, more than 160 data centers were built in 2025 in areas with scarce water resources, with local communities raising concerns that went largely unaddressed. In The Dalles, Oregon, Google’s data centers now account for 25% of the city’s water use. In March, UN-Water released a policy brief specifically addressing the tension between data center expansion and the internationally recognized human right to water.
The Great Salt Lake — already losing about half its surface area to water diversion — sits in the same watershed as the Stratos Project. The developers claim their facility will use “zero water” cooling technology. No independent verification of this claim exists. No manufacturers have been named. No contracts have been published.
Don’t Let Them Make It About Your Prompt
Here’s where I want to be direct.
There is a growing genre of climate and tech commentary that focuses heavily on individual AI users — the carbon footprint of your ChatGPT query, the water cost of generating an image, the electricity burned by your email search. This commentary isn’t wrong. Those costs are real. But there is something worth noticing about how quickly it gets deployed, and by whom.
When the conversation shifts to individual AI consumption, the companies making decisions like the one in Box Elder County can quietly step into the background. The question becomes whether you’re using AI responsibly rather than whether a three-commissioner panel should be able to green-light a 40,000-acre natural gas campus over the objections of 500 constituents.
Your AI prompts didn’t build the Stratos Project. Kevin O’Leary did. The Military Installation Development Authority did. The economic logic of an AI race that treats computation as a strategic resource — the same logic driving the US, China, and the Gulf states to fund massive data center build-outs with minimal public accountability — did.
The beer can is your query. The highway is what they’re building in the desert.
Abbey’s point was not that individual choices are irrelevant — it’s that individual shame is a weapon, and it’s often wielded by the people most responsible for the system producing the problem. We should know what we’re consuming. We should think about how we use AI. But we should not let that thinking become a substitute for looking at the scale of what’s actually being built, by whom, with what consequences, and for whose benefit.
What Gets Silenced When We Don’t
Five hundred people showed up. Three people said yes anyway. And those 500 people will likely lose — not because their concerns weren’t legitimate, but because the forces moving this project forward are operating at a scale and a speed that local democratic processes were not designed to match.
That’s the actual problem. Not that AI is bad. Not that you shouldn’t use it. But that the infrastructure being built to power it is moving faster than accountability can keep up with — and that communities sitting in its path are being told, explicitly or implicitly, that progress doesn’t wait for them.
The human rights question here isn’t abstract. The UN recognized water as a fundamental human right in 2010. Whether the Stratos Project violates that right will be decided by courts and engineers, not county commissioners. But the broader question — who gets to say no to something like this, and whether that no actually counts — is one that democratic societies are going to have to answer, loudly and soon.
Because right now, the answer seems to be: not the 500 people in the fairgrounds.
Takeaway for Teachers
The Box Elder vote is a genuinely useful classroom moment — not because it’s anti-AI, but because it forces the accountability question into the open. Show your students the numbers: 40,000 acres, 9 gigawatts, three votes, 500 protesters. Then ask: who has the power to make decisions like this, and who bears the consequences?
If you want to push further, introduce them to the “carbon footprint” concept — a term popularized by BP in the early 2000s to shift focus from fossil fuel companies to individual consumers. Then ask: what’s the equivalent move happening in conversations about AI and energy? Who benefits when the conversation is about your query instead of their data center?
These aren’t anti-technology questions. They’re citizenship questions. And in a world where AI infrastructure is being built at military speed with minimal public input, knowing how to ask them might be one of the most important things we can teach.
David Jacobson is a high school history teacher. He writes about AI, education, and the messy intersection of the two at shouldiuse.ai.
