Editor's note: This is an opinion piece from the editor of GreeAlityTV — a web applications engineer who has worked with LLMs and AI since they went consumer-grade, and who has spent years running local models and AI services on home servers. It reflects the editor's views.
You Don't Need a Warehouse
Here's something the data center pitch decks won't tell you: everything you're reading right now doesn't need a warehouse. GreeAlityTV runs on a server we built ourselves and host ourselves, right here in Greeley. The software behind it was built in Greeley, too.
The AI tools we use internally run on that same kind of local hardware — equipment that fits in a room, not a campus that drinks a river. They're air-cooled, with no water-hungry cooling systems. And when they're not in use, they shut down automatically, so they aren't burning electricity or generating heat that needs cooling just to sit idle. That's what responsible AI looks like at a local scale.
I say that as someone who works with this technology every day. For simple LLM and AI support — the chat assistants, the writing help, the search, the automation most people actually use — you do not need a hyperscale campus. Consumer-grade models running on local servers handle this work right now, today. I've run them in my own home.
“The idea that everyday AI requires a warehouse-scale facility is a sales pitch, not an engineering requirement.”
Wait — What Is an "LLM," Anyway?
You've probably used one without knowing the name. An LLM — a large language model — is the technology behind chat assistants like ChatGPT, Claude, and the "AI" buttons popping up in your email and phone.
The simplest way to think about it: an LLM is a very well-read autocomplete. It studied an enormous amount of written text and learned the patterns of how language works. When you type a question, it predicts — word by word — what a helpful answer looks like.
Here's the part the sales pitch skips. There are two very different jobs in AI:
- Building the model (called "training") — this is the heavy lifting. It takes massive computing power, and it's done once, up front, by a handful of large companies.
- Using the model (called "running" it) — this is what happens when you ask a question. It's far lighter. Smaller models can run on a single computer, the kind with a good gaming graphics card.
Think of it like an encyclopedia. You need a printing plant to make the book. You don't need a printing plant in every town to read it. Drafting an email, summarizing a document, answering customer questions, searching a website — that's the everyday "reading" side of AI, and it's what most people actually use.
Yes, big tech companies serving millions of people at once do run giant facilities. But that's a business choice to centralize everything in one place — not something a town like ours needs for its everyday AI.
Portfolios Over People
A portfolio is simply a collection of investments — the stocks, buildings, and land that investors, funds, and big companies own to make money. When I say these facilities increasingly serve portfolios over people, I mean they're built first to produce returns for investors, not to serve the people who live next door.
Why the portfolio, and not the people?
Because right now, data centers are one of the hottest investments around. Investors are betting that demand for AI keeps growing — so land with a power hookup, an electricity contract, and access to water becomes a valuable asset. That asset can be held, borrowed against, or sold, much like a stock. The building doesn't have to serve us to be worth a lot to them.
Look at how the math lands locally. The company behind the Weld County project is based in New York and bought its 438-acre site for about $15.6 million. It has promised around 200 long-term jobs. In return, our region commits water, grid capacity, and land for decades. The profits flow to shareholders somewhere else. The noise, the water questions, and the power bills stay here.
It's a bit like an out-of-town landlord buying up a whole block. They may never live there — but the neighbors live with whatever gets built.
And all of this is built on top of the knowledge of the internet, which has been scraped — including the proprietary work of creators, writers, and artists who were never asked and never paid.
Our resources matter. Water. Power. Land. This technology has become a cash-grab position while the middle class struggles to pay a water bill, fuel, and groceries. The incentives of a few beneficiaries are being leveraged, irresponsibly, against our county and city officials — and against what has been a unified, nonpartisan voice of the people.
We Say No
We say no to all large projects — especially on the heels of a record-hot summer and one of the roughest winter and monsoon seasons Colorado has seen. This is not the moment to hand our water and our grid to speculation.
It's never too late to show up. City council meetings. County commissioner meetings. Your voice, in the room, on the record — that's what moves this. Show up and get heard.
Leave a comment below to make your voice heard.
Join the conversation
What do you think? Share your thoughts in the comments below.
Leave a comment ↓Stay in the loop
Get new stories and our newsletter delivered to your inbox.
We'll email you once to confirm. Unsubscribe anytime.