‘AI is going to take away our jobs… AI is ruining your brain…AI is the worst.’
It was my senior fall semester, and the conversation around AI had reached the kind of fever pitch that makes everyone feel like they should have an opinion, and it was typically a negative one. I, on the other hand, had never been exposed to artificial intelligence, so I only thought of it as a glorified search engine. I enrolled in the class to catch up.
Over the course of the semester, I learned what tokens were, how Large Language Models (LLMs) operate, and how many AI models existed to fulfill different needs. AI’s capabilities were hard to grasp even then, and they’ve grown exponentially since.
About a month in, there was an image on the screen that I have not stopped thinking about. It was an aerial photograph of a campus somewhere in the American Southwest: long, low buildings the color of concrete, arranged in rows across a flat expanse of land, surrounded by fences and access roads and not much else. It looked like something between an industrial park and a very serious prison. Then the numbers: Power consumption equivalent to a small city; Millions of gallons of water per day for cooling; Hundreds of acres of land removed from any other use; Communities nearby that would see almost none of the economic activity the facilities generated, but would share all of the infrastructure strain. This was the first time I had seen a hyperscale data center.
The contents of the lecture made one thing clear: Data centers were here to stay.
“Can we learn how to design these more sustainably?”
This question later became the provoking question behind the research that backed my architectural thesis project, “Redefining Urban Infrastructure: Industry Park Fieldhouse,” and has been the engine behind everything I have written since.
I have spent a lot of time learning about the industry, and the clearest thing I can tell you is that the hyperscale data center model has been developed almost entirely without architectural input. Nobody with design training is in the room when most data centers are built. The people making decisions about where these buildings go, how large they are, what they look like, and what relationship they have to the communities around them are not architects or urban designers. They are mechanical engineers and corporate real estate executives, satisfied with the numbers on a proforma and various financial indicators that their investments will be profitable.
The result is a building optimized for things that have nothing to do with the places where these buildings land. A building type that looks the same in Virginia as it does in Arizona, that could be placed anywhere on a flat piece of land with adequate power and water, and that has been deliberately designed to communicate nothing — no civic presence, no acknowledgment of its surroundings, no admission that people live nearby.
However, there are data centers in the world right now that contribute to the energy grids of the buildings around them, that occupy urban land responsibly, that have been designed to be honest about what they are and what they cost. They are exceptions, and they exist for one reason: somewhere in the process, someone with design training was in the room, and insisted the building owed something to the place it landed in.
My architecture training taught me to think about buildings as the mediators between human beings and the world they build together — not just as technical objects to be optimized, but as civic acts. A building should be the result of holistic design: a reflection about what a community values, what it owes its members, and how it imagines its future. It is a vital instrument in the orchestra we call civic life.
The concern with data center development today is a lack of holistic design, and therefore holistic impact. The building type that houses the entire infrastructure of artificial intelligence has been developed without consideration of the human. We cannot meaningfully assess our impact as designers of the human experience and as stewards of the built environment without taking the human into account.
So, can we learn how to design data centers more sustainably?
Sustainable design is not as easy as pointing at a building and saying “but what if it were better.” There are real constraints — financial, physical, operational — and any honest engagement with this building type has to reckon with them. But the constraints have been used as an excuse to avoid the question entirely, and that avoidance has a cost that is accumulating in real places, on real land, in real communities.
Developers build what the market accepts and regulators permit. For now, the market accepts windowless boxes and regulators have not asked for anything more. That can change, but only if the people who think about buildings as civic acts decide the question is worth asking out loud.
The infrastructure of artificial intelligence is being built right now, without anyone in the room whose job is to think about the human standing next to it.
This series is my attempt to be in the room, one article at a time.