AI Data Center Compute Capacity, U.S. Versus The Rest Of The World
News this week suggests U.S. hyperscalers are doubling down on AI. Google raised its projected 2026 capital spending to $195 to $205 billion (from $180 to $190 billion previously). But will the revenue materialize to justify the capex spend? Well, in Q2 2026, Google Cloud revenue surged 82% year-over-year, while its backlog nearly quadrupled during the same period, meaning demand already exceeds available capacity. It’s no surprise then that U.S. data center capacity more than tripled in the last 12 months, with America now home to 93% of global AI compute capacity. That said, the AI boom is more global than it appears. Almost all of U.S. high-tech equipment is imported, driving demand for AI hardware across Asia, while foreign customers pay American hyperscalers to run AI workloads on U.S. infrastructure. However, as AI compute demand shifts from training to inference, developing a good model is not enough: one must also provide the compute necessary to run it in close geographic proximity to the end user. The U.S. AI buildout isn’t just getting bigger—it’s driven by accelerating demand and revenue growth. And, if AI is a race to deploy intelligence at scale, then the U.S. is the clear leader.
