Trillions Won't Fix the AI Data-Center Bottleneck

Despite forecasts of roughly $1 trillion in AI-data-center spending, cash isn’t the bottleneck. The big limits are power availability, skilled labor, land, and permitting, plus regulatory pushback on new builds. BloombergNEF warns of a 19 GW power shortfall by 2035 if growth persists, and utilities may approve only a fraction of requested hookups. As a result, AI data-center deployment is likely to proceed more slowly and in fits and starts, even as demand remains robust and suppliers of GPUs, CPUs, memory, and power equipment stand to benefit. In a worst-case scenario, compute could be constrained and users shift to cheaper models, but the current trend is toward continued heavy investment despite the bottlenecks.