The rapid expansion of artificial intelligence depends on a massive, physical architecture: the data center. But Silicon Valley’s ambition is increasingly colliding with the stubborn realities of land, labor, and the electrical grid. According to an analysis of satellite imagery from geospatial firm SynMax and permit data from IIR Energy, nearly 40 percent of U.S. data center projects are likely to miss their scheduled completion dates this year.
The delays, often exceeding three months, affect major builds for industry titans including Microsoft, Oracle, and OpenAI. While the capital being poured into these "gigascale" projects is unprecedented, the infrastructure required to support them cannot be summoned as quickly as code. Satellite views of construction sites show significant lags in clearing land and laying foundations, suggesting that the digital future is being throttled by a lack of physical progress.
The bottlenecks are multifaceted. Industry executives cite chronic shortages of specialized tradespeople, such as electricians and pipe fitters, alongside a scarcity of essential power equipment. Beyond the job site, securing the massive amounts of electricity required—often equivalent to the needs of hundreds of thousands of homes—has led to protracted permitting battles and growing local resistance. As the industry attempts to build its way into the next era of computing, it is finding that the most significant hurdles are not algorithmic, but industrial.
With reporting from Ars Technica.
Source · Ars Technica
