Not the Datacenters, How We Build Them

I’m a fan of AI. I don’t dispute that, and I won’t pretend otherwise. I use it every day, and part of my income comes from helping other people use it. I’m about as biased toward it as a person can get.

I don’t even mind that it needs huge datacenters, or that those datacenters demand enormous amounts of power and water. That’s a real cost, and it’s one I’m mostly willing to accept.

What I mind - what I mind very much - is how these datacenters are being rolled out. They’re being forced onto communities without regard for those communities, or for their impact on the rest of the world in general. And when people push back, they get dismissed. “Get used to it.” “AI is the future.” As if that settles the matter.

The way the build-out is going right now is simple to describe: it maximizes the benefit to the tech companies and does it to the detriment of everyone else. And everyone else is noticing.

Why Density Makes Sense

The best way to provision a datacenter is to make it as dense as possible. The closer the machines are to one another, and the more of them you can fit in one place, the better off you are.

Power doesn’t have to travel as far. Cooling works more efficiently. Data has less distance to cross, so everything runs faster. Maintenance and upgrades are far easier when all the hardware is in one building instead of a dozen. Density is cheaper per unit of compute, and it’s more compute per dollar.

There’s no arguing with any of that. Density is the right way to build, period.

What Density Costs Everyone Else

But density has a physical shape, and that shape has to go somewhere. The result is predictable. Amazon is building one right now in Pecos County, Texas, and it shows exactly what I mean.

Eight thousand acres, three buildings, and 35 natural gas turbines good for around 7.65 gigawatts. If it’s finished as permitted, it’ll be the largest power plant in American history, bigger than the Grand Coulee Dam.

Start with the turbines. They’re loud, and they’re heavy polluters. This one isn’t even drawing off the grid. It’s “behind the meter,” generating its own power instead of waiting years for a grid connection, so Amazon can honestly say it won’t raise your electric bill. It won’t. What it will do is put out up to 33 million tons of carbon dioxide a year.

Back of the napkin: a typical car puts out about 4.6 tons of CO2 a year. Divide it out, and that one plant is the same as putting another 7 million cars on the road. Not spread across the country. Parked next to Fort Stockton. That’s before you count the 12,000 tons a year of soot, ammonia, and carbon monoxide its permits allow.

Then there’s the cooling. Evaporative, which is loud and drinks water. Amazon says it’ll use brackish groundwater, not the water people drink or irrigate with. That’s the exception they always reach for.

And the land. Eight thousand acres, though to be fair this one’s desert. Nothing out there. So it’s not farmland they’re paving over here. That’s just this site. But a lot of other datacenters do take farmland - Iowa cornfields, Ohio hay fields - flat and cheap and growing food, and you don’t get that back.

The people who live there have already noticed. “Who decided this?” one of them asked the local news. “Did they even think about us?” Another put it better than I can: without regulation, it’s, in his words, “like forcing everybody to start smoking cigarettes.”

So no, they didn’t drain the grid and raise your bill. They built their own turbines instead. Same cost, just moved: it lands on your lungs instead of your electric bill.

The Alternatives Already Exist

This is the part where it gets frustrating. Every reasonable alternative is already on the table. They’re just less efficient - for the tech companies.

Compute in the Closet

Distribute the compute into people’s homes. This is the one that gets dismissed fastest, and I think it’s the one normal people would like best.

Think about what a house already is. It’s standing. It’s wired for power and internet. It’s heated and cooled and paid for. Now imagine a box in a closet doing some of the work that would otherwise happen in a field outside Fort Stockton, and in the winter it dumps its waste heat into your house instead of the sky.

The company covers the box, the power, the climate-control (AC and/or heating), the bandwidth, and they pay you a little every month for hosting it. You don’t have to understand what it’s doing. It sits there, warms the spare room, and knocks something off your mortgage.

Which would you rather have next door: a 35-turbine gas plant, or a few neighbors each running a small box that pays their bills?

The honest reason it never happens is the same as everything else in this essay. It’s not efficient. For them. A million little boxes in a million homes is a coordination nightmare next to one campus they fully control, so they build the campus. The inefficiency lands on their side. We’d be fine either way.

Solar and Wind

Build out solar and wind instead. I did some back-of-the-napkin math on that same Pecos plant and came out to something like 120,000 acres of solar, not counting the batteries. My first thought was: what’s wrong with that? It has an environmental impact, sure. But is it worse than the gas turbines they’re putting in instead? Wouldn’t it be easier for communities to accept if all the noise was isolated away behind a few thousand acres of solar panels?

Pay the Neighbors

Pay the people who have to live with it. If you’re going to build a loud datacenter next to someone, pay them for it. A large solar farm with the surplus sent to the community, a water treatment plant, even a whole sewer system. High-speed internet, since they have to build the connections anyway. Or straight cash. $100K a year to a household can go a long way.

More, Smaller Datacenters

Build more, smaller datacenters. They’re already all over the DFW area - I’ve worked in them. All the banks have them - BofA, Capital One, JPMC…all of them. There are quite a few smaller ones all over downtown, up and down US75. They’re pretty much everywhere already. That model can be extended: smaller facilities spread around cities and towns rather than one giant one dropped on somebody’s county line.

But, why not?

The only real problem with any of those is that none of them are as efficient for the tech companies as a single huge datacenter running on gas turbines and cooled with evaporative air.

But none of that extra expense is actually a burden to them. OpenAI and Anthropic are both planning multi-billion-dollar IPOs, and they’re expected to be worth a trillion dollars. Compared to numbers like that, the cost of doing this the right way barely registers. They don’t run gas turbines because the alternatives are too expensive. They run them because they can, and because nobody makes them do otherwise.

The Part People Are Pissed About

There’s a lot of talk about how AI is supposed to benefit everyone, but all the benefit is flowing to trillion-dollar tech companies while the cost lands on everyone else’s doorstep. Based on the PR push I’ve been seeing from Amazon and others, they expect some advertising to change peoples’ minds. It’s all too obvious and egregious to be pushed away that easily.

I know there’s no chance in hell that anybody who can actually make a decision about this will read these words, much less be swayed by them.

I’m getting pissed, too…and I like AI.