The AI arms race just ran headfirst into a problem Silicon Valley can't code its way out of: people who actually live places.

The Summary

The Signal

Box Elder County just greenlit something absurd. The Stratos Project would sprawl across 40,000 acres in Hansel Valley, more than twice the size of Manhattan. It would pull 9GW of power in a state where peak demand currently hits around 5GW. The pitch: American AI dominance. The reality: a venture capitalist from Shark Tank wants to build the world's largest data center in a desert valley where water is already contested.

The opposition isn't fringe. Experts warned commissioners. Locals showed up angry. They approved it anyway. This is the new infrastructure conflict: AI companies need massive compute, cheap land, and reliable power. Rural counties need economic lifelines. The math looks simple until you add water tables and electricity grids that weren't built for this.

"The only solution that will work is simply paying them off."

Stratechery cuts through the posturing: people oppose data centers for understandable reasons. These facilities bring construction jobs that disappear, permanent jobs that don't employ many locals, and infrastructure strain that lasts decades. The usual playbook, promising tax revenue and economic development, doesn't compensate the people who actually bear the costs. Higher electricity rates. Depleted aquifers. Traffic during construction. The benefits accrue to county budgets and distant shareholders. The costs hit households.

The fix isn't complicated, just uncomfortable for companies used to extracting value without direct payment. Write checks to residents. Not to the county. To the people. If a data center makes economic sense, the operator can afford to make locals financially whole. If it doesn't pencil out after fair compensation, it shouldn't get built.

Key tensions in play:

  • National AI competitiveness vs. local resource constraints
  • Corporate profit extraction vs. community compensation
  • Infrastructure built for one era straining under demands of another

This is playing out everywhere. Virginia's Loudoun County already hosts the world's highest concentration of data centers. Texas is building them across the Permian Basin. Ireland capped new data center development in Dublin because the grid couldn't handle more load. Utah's Stratos Project is just the largest, most brazen version of a pattern: AI infrastructure needs colliding with physical reality and local political will.

The agent economy runs on compute. Compute runs on electricity and cooling. Both require resources that someone, somewhere, already depends on. The companies building Web4 are discovering that "move fast and break things" doesn't work when the things you're breaking are water rights and power grids in places where people have been living for generations.

The Implication

If you're building AI infrastructure, budget for direct community payments from day one. Not PR campaigns. Not job training programs that never quite materialize. Actual cash to actual people whose lives you're about to change. The math is straightforward: if your margins can't handle fair compensation for externalized costs, your project doesn't actually pencil out.

Watch how Utah's Stratos saga unfolds. If it proceeds without direct resident compensation and infrastructure collapses or political backlash kills it mid-construction, that's the roadmap for what doesn't work. If other counties see this deal and start demanding upfront payments to households, that's the new baseline. The age of building massive infrastructure projects on vague promises is ending. The age of paying people directly for what you're taking is just beginning.

Sources

The Verge AI | Stratechery