AI
“According to the Financial Times citing regulatory intelligence outfit RegLayer, 47 large data center projects have already fallen through in 2025.”
Plain restatementThe Financial Times, using data from a firm called RegLayer, reported that 47 large data center projects were cancelled during calendar year 2025.
The Financial Times did report this figure, and the firm RegLayer is real, but two details in the claim are wrong. The FT published the number on 18 October 2026, and it covers 2026 so far, not 2025. The FT also said at least 47 large US projects were "delayed, withdrawn or blocked," which is not the same as 47 projects falling through. A delayed project is still going ahead, and RegLayer says its own count is conservative because it only tracks the biggest sites. Adding to the confusion, a different organization reported that at least 48 US data center projects were blocked or stalled by local opposition during 2025, which is a separate count from a separate source. What cannot be checked is whether the number itself is accurate, because RegLayer has not published the list of projects or the rules it used to count them. The wider point in the reporting, that US data center plans are being slowed by financing costs, permits and local opposition, is what the FT article describes.
[drifted from the evidence:] According to the Financial Times [drifted from the evidence:] citing regulatory intelligence outfit RegLayer, 47 large data center projects [drifted from the evidence:] have already fallen through in 2025.
The Financial Times, [added by the neutral restatement:] using data from a firm called RegLayer, [added by the neutral restatement:] reported that 47 large data center projects [added by the neutral restatement:] were cancelled during calendar year 2025.
Red-tinted words in the claim drifted from the evidence. Green-tinted words are what a neutral restatement needs.
The trace / claim to source
- The Financial Times did report this, in an article published 18 October 2026, and did cite RegLayer as the data source.
- RegLayer is a real firm. It operates a public website describing itself as a regulatory intelligence provider for technology companies.
- The number 47 is the number the FT reported.
- The related figure in the surrounding post, that Bernstein assessed only about 35 percent of the planned US data centre pipeline as credible, also appears in the same FT reporting.
- The broader picture the post describes, that US data center projects are running into financing costs, permitting and local opposition, is what the FT article reports.
- The claim places the figure in 2025. The FT reported it for 2026, covering the year to date as of 18 October 2026. The post's own caption said "this year," which was correct at publication; the version of the claim circulating here converts that to 2025, which points the reader at a different year. This matters because a separate tracker reported a similar-sounding figure, at least 48 projects blocked or stalled by local opposition, for calendar 2025. The two are different organizations measuring different things, and treating the 2026 count as a 2025 count merges them.
- "fallen through" means cancelled. The FT reported projects "delayed, withdrawn or blocked." A delayed project is still alive, and a blocked project may be revived or relocated. Only the withdrawn category clearly matches "fallen through," and the breakdown among the three categories was not published. The claim converts a mixed-status count into a cancellation count.
- The reported figure is "at least 47," and RegLayer describes its own count as conservative because it tracks only the biggest sites. Dropping "at least" presents a floor as a total.
- The count covers US projects. The claim states it without geographic scope, which reads as global.
- Rumor as fact is not the issue here, but single-chain repetition is: dozens of outlets carried this number within a week, all sourced to the one FT article. That breadth of coverage is not corroboration of the count.
- Whether the 47 figure is accurate as a measurement cannot be assessed. RegLayer's inclusion criteria, project list and methodology are not public, so the count cannot be independently audited or reproduced.
- The split between delayed, withdrawn and blocked within the 47 was not published, so how many projects genuinely ended is unknown.
- Whether RegLayer has published any comparable count for calendar 2025 could not be established. No such figure was found.
- The current figure may already be higher, since the count was reported as year to date as of 18 October 2026 and RegLayer called it conservative.
- Secondary claims in the same post not investigated here: the CoreWeave 27 percent share price decline, and the characterization that Anthropic and OpenAI "deemphasized due diligence."
The Financial Times published "America's $10tn AI infrastructure boom is starting to fray" on 18 October 2026. The FT's own social accounts carry the piece with that date. RegLayer is a real company that sells regulatory intelligence to technology firms. The figure the FT reported is for 2026, not 2025, and the status is not cancellation. The FT sentence as reproduced on a public forum reads that "At least 47 large projects have been delayed, withdrawn or blocked in 2026," attributed to data shared by RegLayer. Every independent report of the FT piece found, from Techmeme, Heise, Heatmap, Seoul Economic Daily, Dealroom and others, states the same year and the same three-part status wording. Several add that RegLayer itself calls the count conservative, because it tracks only the largest sites. The other figures in the surrounding post also trace to the same FT article: Bernstein estimating that only about 35 percent of the planned US data centre pipeline is credible, and the quoted remarks from unnamed executives. A separate and different statistic exists for calendar 2025: a tracker called Data Center Watch reported at least 48 US projects blocked or stalled by local opposition during 2025. That is a different organization, a different count, and a narrower cause category.
Complete reasoning
The reply is formatted for pasting into the thread where the claim is circulating.
Ask this case
Answers come only from the case file above; nothing is added.
Did the Financial Times really report that 47 data center projects fell through?
The FT did cite a figure of at least 47 large US data center projects from a firm called RegLayer, but it did not call them cancelled. The FT described them as delayed, withdrawn or blocked, which is a broader category than projects falling through.
Is the 47 figure for 2025 or 2026?
It is for 2026. The FT article was published on 18 October 2026 and the figure covers that year to date, not 2025 as the claim states.
What does delayed, withdrawn or blocked actually mean?
These are three different outcomes lumped into one count. A delayed project is still moving forward and a blocked one could still be revived or relocated, so only withdrawn projects clearly match the idea of a project falling through. The breakdown between the three categories was not published.
Is there a separate figure for 2025 that people might be confusing this with?
Yes. A different tracker called Data Center Watch reported at least 48 US data center projects blocked or stalled by local opposition during calendar 2025. That is a separate organization measuring a separate thing from RegLayer's 2026 count.
Can the 47 number itself be verified as accurate?
The investigation could not establish this. RegLayer has not published its project list or its methodology, so the count cannot be independently checked, and RegLayer itself says the figure is conservative since it only tracks the largest sites.