Two trackers studying setbacks in America’s data-centre boom are reporting sharply different counts, exposing a gap between headline project totals and the information investors need to assess whether announced AI computing capacity will actually be built.
RegLayer identified at least 47 major US data-centre projects delayed, withdrawn, or blocked during 2026, according to a Financial Times investigation. Meanwhile, Data Center Watch, a project of research firm 10a Labs, recorded at least 75 projects disrupted by local opposition in the first quarter alone.
Those developments carried approximately $130 billion in proposed investment value.
RegLayer’s count, as reported by the FT, concerns major developments. Data Center Watch focuses on projects affected by community opposition, including permitting challenges, moratoriums, and withdrawals.
Data Center Watch also distinguishes between projects blocked and projects delayed. The former includes proposals withdrawn following opposition, while the latter can remain active despite missing their original timelines.
The group’s first-quarter findings were already public in June. Its second-quarter report identified 45 projects worth nearly $68 billion blocked or delayed between April and June, down from 75 in the preceding quarter.
The second-quarter report also cited continuing opposition to projects previously disrupted, making it inappropriate to treat the quarterly counts as 120 unique failed developments.
The numbers instead point to a more consequential distinction between a proposed facility and one capable of supplying paying customers with computing capacity.
Investment forecasts can include projects at substantially different stages, from preliminary announcements to permitted construction and completed facilities awaiting electricity. A setback recorded by an opposition tracker does not, by itself, establish how much computing capacity will be unavailable, for how long, or whether customers have contractual remedies.
Bernstein Research estimates that approximately 35% of the planned US data-centre pipeline is credible, according to the FT.
Electricity regulators are attempting to establish more consistent information on one part of the development pipeline. On October 9, the Federal Energy Regulatory Commission approved a large-load registry proposed by PJM Interconnection, which operates the grid serving much of the Mid-Atlantic and Midwest.
The registry covers large electricity users meeting a 50-megawatt threshold and will gather information including expected service dates and power requirements. FERC rejected separate proposed provisions governing service and curtailment.
The registry will not reconcile the competing national project tallies. That would require comparable project identities, development stages, power-connection dates, and measures of planned capacity.
Until those details are available, the 47 and 75 counts document disruptions but establish neither the share of projects that will fail nor how much of America’s advertised AI infrastructure pipeline will reach operation.