I find myself thinking about this story from Shelbyville, Indiana more than I probably should. A mayor caught on camera dismissing residents who don’t want a $2 billion data center in their neighborhood, calling their homes “shitty houses” and labeling them as rentals. The whole thing is ugly, and frankly, it speaks to something much bigger than one politician’s poor choice of words.
Here’s why I keep coming back to it as a developer.
The data center being proposed isn’t some generic commercial development. It’s infrastructure. Real, physical infrastructure that powers the artificial intelligence systems we building, deploying, and increasingly depending on. Every model we train, every inference we run, every API call to an AI endpoint somewhere, ultimately depends on physical buildings full of servers consuming enormous amounts of electricity and water.
That’s the backdrop for what’s happening in Shelbyville, and it’s happening in dozens of other communities across the country right now. Data center projects are facing pushback everywhere from Virginia to Arizona to right here in Indiana. And the uncomfortable truth is that we, as an industry, haven’t figured out how to have honest conversations about what this infrastructure actually costs.
The scale problem nobody talks about
Let me break down what’s actually happening with AI infrastructure demand. Modern large language models and the systems built on top of them require computational resources that dwarf traditional software workloads. We’re talking about training runs that consume as much electricity as small cities, cooling systems that evaporate millions of gallons of water, and facilities that need to be located somewhere.
The mayor’s comments in that video reveal a troubling disconnect between the people building this future and the people who have to live with its physical manifestations. When you’re a developer working on the 47th layer of abstraction away from the hardware, it’s easy to forget that every API call has a physical footprint. But communities don’t have that luxury. They see the power lines, they feel the traffic during construction, and they worry about their property values and their water.
I’ve been in enough tech conferences to know that this reality rarely comes up in talks about scaling AI systems. We’re too busy discussing inference optimization and model distillation to think about the fact that somewhere, someone is fighting to keep a data center out of their neighborhood.
This isn’t just a local problem
The interesting thing about the Shelbyville situation is how quickly it became about more than just local zoning. The mayor’s dismissive language toward working-class residents, particularly renters, touched a nerve that resonates far beyond Indiana. We’re in a housing crisis. People are struggling to afford homes. And here’s a mayor essentially telling people that their homes don’t matter because they don’t own them.
That’s the human layer that we can’t abstract away, no matter how elegant our software architectures are.
For developers working in artificial intelligence, this should matter because the solutions we’re building will only work if we maintain public trust. Every time something like this happens, it erodes a little more of the social license that allows our industry to operate at this scale. We can build the most sophisticated models in the world, but if communities keep pushing back against the infrastructure needed to run them, we’re going to hit a wall.
The data center debate isn’t separate from AI development. It’s fundamentally part of it. When a city council votes down a proposed facility, that’s not just a local zoning decision, it’s a constraint on computational capacity for the entire region. These decisions have downstream effects on what we can build and where we can deploy it.
What we can do differently
I’m not naive enough to think that developer sentiment alone will change anything. The economics of AI infrastructure are what they are, and the demand isn’t going away. But I do think we need to start incorporating this reality into how we think about system design and deployment strategies.
There’s been a lot of talk about edge computing and distributing inference loads more broadly. Maybe it’s time to take those conversations seriously. If we can reduce the concentration of computational needs in massive data centers, we might be able to avoid some of these community conflicts entirely. Edge deployment means infrastructure that’s closer to where people live, but in smaller, more manageable increments.
We should also be thinking about what it means to build AI systems that are more efficient. Not just computationally efficient in the abstract sense, but efficient in terms of real-world resources. Every FLOPS we can save is electricity that doesn’t need to be generated, cooling that doesn’t need water, and infrastructure that doesn’t need to be built in someone’s backyard.
The industry needs advocates who can speak honestly about these tradeoffs. We need people who can sit in a room with community members and explain what a data center actually means for their neighborhood, not just in terms of jobs and tax revenue, but in terms of noise and traffic and visual impact. That’s not easy work, and frankly, it’s not the kind of work that gets you promoted at most tech companies.
But maybe it should be.
The Shelbyville mayor said the quiet part out loud, and now everyone is dealing with the fallout. What happens next in that small Indiana city will probably determine whether the data center gets built, but the wider implications extend far beyond one project. We’re at a point where the physical realities of AI infrastructure are catching up with our ambitions, and how we handle that tension will shape what kind of technological future we actually get to build.