A 43-year-old rancher named Kassi Solberg has been driving into Broadview, Montana, to sit through city council meetings and ask a fairly basic question: What exactly is being built next to her property, and who decided that? Broadview has something like 140 residents, a single bar and restaurant, and a school gymnasium that doubles as the Friday night social calendar.
The project she keeps asking about is a 5,000-acre artificial intelligence campus proposed by a young Texas company called Quantica Infrastructure, a footprint roughly the size of 3,800 football fields that would eventually draw up to a gigawatt of power. To put that number in context, a single gigawatt is more electricity than the local utility currently supplies to its entire base of customers across the state
I am not writing this to make Solberg a hero or Quantica a villain. The interesting part of this story is not the cast. It is the collision.
For the last three years, the AI conversation has lived almost entirely inside software. We have argued about model releases and benchmark scores, about GPU allocations and funding rounds, about whether agents are going to rewrite how we build applications and run companies. I have written plenty of that myself. What the Montana story makes plain is that none of it happens in the cloud in any literal sense.
At some point, the abstraction ends and the thing has to be built somewhere, on actual land, drawing actual water, pulling actual current off a grid that somebody has to pay to expand. AI stopped being only a software story a while ago. It is now a power story, a land story, a water story, a transmission story, and increasingly a local politics story.
The piece the AI datacenter industry keeps getting wrong, and I say this as someone who has watched a few of these waves arrive and recede, is the nature of infrastructure itself. Technology people are conditioned by software economics, where the marginal cost of the next user rounds to zero and you can scale across the planet in a quarter. Infrastructure has never behaved that way. A power plant takes years. A transmission line takes years, and in Broadview, the cost of a roughly 21-mile line has already become a fight over who pays, the company or the ratepayers. Permitting takes years. Earning the trust of a community that was never consulted takes longer than all of it combined. Physics and county commissions do not care about anyone’s product roadmap. I watched the telecom industry learn this during the fiber buildout, and watched broadband providers relearn it a decade later when they discovered that laying cable through someone’s town is a negotiation, not a deployment.
There is also a new economic equation underneath all of this that communities are right to scrutinize. For a century, towns swallowed hard about noise, traffic and pollution because the factory or the plant employed half the county for two generations. The deal was legible. A data center is a different animal. It can represent billions in capital and enormous draws on power and water while supporting a permanent workforce you could seat in the town’s only restaurant on a slow night. Some locals see opportunity in that, and they are not wrong to; the man who owns that restaurant would happily take the foot traffic. But the old social contract, where heavy industry bought its welcome with payroll, does not transfer cleanly to a building full of servers and a handful of technicians.
The reflex in our industry is to file all of this under anti-technology sentiment, to wave it off as NIMBYism from people who do not understand the future. That is a lazy read and a costly one. The people showing up to these meetings are not debating transformer architectures or AGI timelines. They are asking about wells and aquifers, about whose electric bill goes up, about road wear and tax abatements and what happens to property values and the night sky. Those are governance questions, accountability questions, and questions about local control. When a company representative tells residents they have no role in the conversation, or a county waves a project through because that particular stretch of land happens to carry no zoning, the resulting anger is not technophobia. It is the predictable response to being told that something enormous is going to happen to you, and your job is to get out of the way.
Step back from Broadview, and the same pattern is forming nearly everywhere, which is exactly why this small story matters well beyond Montana. The working assumption for years was that the AI race would be won by whoever had the smartest researchers and the most advanced chips. That is still true, but it is no longer sufficient. Access to electricity has quietly become a competitive variable on par with talent and silicon. Utilities across the country are tearing up demand forecasts they wrote two years ago. Developers are floating dedicated natural gas plants to feed individual campuses. Nuclear, including the small modular reactors that were a punchline not long ago, is back in serious conversation, and hyperscalers are signing power deals that would have sounded absurd in 2021. What is taking shape is one of the largest infrastructure buildouts in the history of modern technology, and it is moving faster than the permitting, the grid, or the affected communities can absorb
None of this means the data centers should not get built. The compute has to live somewhere, and rural communities with open land, fiber and proximity to power are logical hosts. It does mean the industry would be wise to treat places like Broadview as partners in a long relationship rather than obstacles to route around. The companies that win the infrastructure phase of AI will not necessarily be the ones that move fastest. They will be the ones that show up with real numbers, share the cost of the transmission line, tell the truth about water, and understand that a signed land deal is the beginning of the conversation, not the end of it.
The AI industry spent the last several years worried almost entirely about models, convinced that the hard problems all sat on the other side of a benchmark. It is about to spend the next several years discovering that power, infrastructure and the trust of a rancher in Montana are every bit as hard, and every bit as decisive. The frontier moved while we were all watching the leaderboard. It runs through Broadview now.
