US data centers pursuing power for the artificial intelligence boom are unlikely to secure more than two-thirds of the electricity they seek, according to reporting that cites “phantom” projects and low-probability proposals. The gap points to a mismatch between announced demand and what ultimately gets built and connected to the grid.
Both outlets describe how many of the initiatives remain uncertain, with some plans not progressing beyond early filings, while others face delays or do not advance past preliminary planning. The reporting characterizes a portion of the pipeline as long-shot pitches—proposals that may never reach construction or commissioning—and highlights the resulting risk that grid operators and power providers will not see the full amount of capacity developers claim they need.
While the articles align on the central finding—an overhang of requested electricity that is unlikely to materialize—they frame it in the context of differing data and market signals, using similar language about projects that do not translate into real-world demand.
