Satellite imagery shows 40% of US AI data centers will miss 2026 deadlines

Geospatial analysis from SynMax, cross-referenced with permit filings tracked by IIR Energy, indicates that nearly 40 percent of US data center projects scheduled for completion this year are running more than three months behind. Sites tied to Microsoft, Oracle, and OpenAI show insufficient ground clearing and foundation work to meet stated timelines, undercutting the aggressive buildout narrative coming out of Silicon Valley.

The bottlenecks are structural rather than financial. Industry executives point to shortages of electricians, pipe fitters, and other skilled trades being stretched across too many concurrent megaprojects, alongside permitting friction and equipment scarcity. Tariffs on Chinese transformers have tightened an already constrained supply of grid hardware.

Power remains the harder ceiling. Utilities cannot spin up generation or transmission capacity fast enough to feed facilities that each draw the equivalent of hundreds of thousands of homes, and local opposition to siting is intensifying. The compounding effect is that compute capacity for frontier AI workloads is slipping out of phase with the capital already committed to it.