The Great Unbuilding of AI
The AI boom is hitting a wall. For the first time ever, US data center construction is declining. The bottleneck isn't capital; it's finding a place to build.
Note: A generated audio podcast of this episode is included below for paid subscribers.
⚡ The Signal
The AI gold rush has an infrastructure problem. Demand for compute is vertical, fueled by an insatiable need for more powerful chips and models. Yet the physical backbone of this revolution—the data center—is hitting a wall. For the first time ever, data center construction in the US is declining not from lack of capital, but from a shortage of viable places to build.
🚧 The Problem
Finding a home for a new data center has become a high-stakes, painfully slow gamble. The process is a minefield of hyper-local challenges: arcane zoning laws, exhausted power grids, and fierce community pushback over noise, water, and energy consumption.
Hyperscalers and developers spend millions on due diligence, navigating a fragmented landscape of utility reports, town hall minutes, and environmental assessments. This manual, consultant-heavy approach takes months, often ending in costly dead ends. The friction between the digital world's exponential growth and the physical world's regulatory inertia has created a critical bottleneck.
🚀 The Solution
Enter Locus, an intelligence platform for data center site selection. Locus replaces months of manual research with a data-driven dashboard, allowing developers to find and validate optimal sites in weeks.
By ingesting and analyzing thousands of disparate data sources—from real-time grid capacity and zoning regulations to public sentiment from community planning documents—Locus identifies sites where power, permits, and people are in alignment. It de-risks multi-billion dollar capital investments by flagging the hidden hurdles before the first dollar is spent on engineering or legal.
