Pundit ClaimsAI claims by Gary Marcus and Ed Zitron, with the evidence

“While the Anthropic-Musk compute deal is an obvious sign about xAI’s lack of demand for compute, it’s also, as I mentioned earlier, a clear sign that AI data centers are mostly not getting finished, and those that do get finished are taking two or three years even for smaller builds.”

Where Are All The Data Centers?, paragraph 148. The full statement is at the source.

Our reading, the sentence that was checked:

The Anthropic–xAI compute deal shows that most AI data centers are not getting finished, and those that are take two to three years even for smaller builds.

Assessment: Unresolved Evidence as of

Where and when Ed Zitron wrote it

Where's Your Ed At, Where Are All The Data Centers?, paragraph 148.

Said
Source observed
live page, full text, public
Evidence as of
the cutoff for this assessment
Assessment recorded
assessment wyea-where-are-all-the-data-centers-5e40a7-a1

We recorded 25 claims from this post. See every claim from this post.

The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2026-05-12. This page does not claim the wording was captured then.

What the claim commits to

Kind
Observation
Deadline, as the speaker wrote it
None stated. This site does not add one.
Deadline state when assessed
No horizon
Deadline state on 2026-10-29
No horizon (the release’s reference date)

From the speaker

  • the Anthropic-Musk compute deal
  • AI data centers are mostly not getting finished
  • taking two or three years even for smaller builds

Our interpretation

  • Treat this as an observation about construction outcomes. The deal is offered as an indicator, not as a cause of construction delays.
  • Use January–March 2026, the latest completed calendar quarter before publication, to assess the ongoing industry state.
  • Read 'mostly not getting finished' as a majority of projects failing to reach usable operation when completion could reasonably be expected. Counting recently announced projects whose completion dates remain in the future would not establish failure.
  • Read 'two or three years' as a typical construction-start-to-usable-operation duration of approximately 24–36 months, including smaller facilities. Isolated exceptions would not settle this industry-wide claim.

Counts as supported if: A representative AI-project cohort shows that more than half of projects expected to be operational remain unfinished, and completed smaller projects typically require approximately 24–36 months from construction start to usable operation.

Counts as contradicted if: Comparable representative records show that most projects reach usable operation as expected, or that completed smaller facilities typically take materially less than two years. An individual fast build is insufficient.

Retrospective: these criteria were written after the outcome was public.

Current assessment

Was it true for the period described?

Outcome: Unresolved

The record was checked, and the evidence available by the cutoff does not settle it.

Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.

Evidence as of
Deadline state when assessed
No horizon
Assessment recorded
Policy version
2026-10-28.7

Rationale

The records show construction risks, but do not establish both parts of the claim. Sightline's broader dataset identifies an unstarted pipeline while giving typical build times of 12–18 months. Those figures neither establish an AI-only majority failing to finish nor settle durations for smaller AI facilities. The Anthropic agreement establishes access to existing capacity, not the industry's completion rate.

Limitations

  • Results depend on the project denominator and the meaning of completion.
  • Available industry research mixes facility types and observed progress with forecasts.

Evidence (3 items)

  • Context only: Sightline Climate, now Currence,

    typical build timelines of 12–18 months.

    Why it matters here: The research identifies a substantial unstarted pipeline and gives a shorter typical construction timeline. It covers large data centers and AI factories together, and future scheduled deliveries cannot establish that most AI projects had already failed to finish.

    Publisher
    Sightline Climate, now Currence
    Kind of source
    Independent measurement
    Describes events of
    Public from
  • Context only: Epoch AI,

    We assume a default 150-day lag between building roof completion and this milestone, which reflects the typical time required for interior fit-out, rack installation, and network integration.

    Why it matters here: The methodology shows why an estimated first-power date is not necessarily a verified completion date. This earlier methodological publication does not measure the target quarter's industry completion rate.

    Publisher
    Epoch AI
    Kind of source
    Independent measurement
    Describes events of
    Public from
  • Context only: Anthropic,

    all of the compute capacity

    Why it matters here: This confirms the transaction invoked by the speaker. A transfer of access to an existing facility does not measure construction completion rates or build times, and occurred after the target quarter.

    Publisher
    Anthropic
    Kind of source
    Company disclosure: the company describing itself, not an independent finding
    Describes events of
    Public from

What would settle it

A representative project register with original schedules, construction starts, verified usable-operation dates, and size categories for AI facilities.

Where it was said (1)

Said once in the sources this register covers.

  • : Where's Your Ed At, Where Are All The Data Centers?

    While the Anthropic-Musk compute deal is an obvious sign about xAI’s lack of demand for compute, it’s also, as I mentioned earlier, a clear sign that AI data centers are mostly not getting finished, and those that do get finished are taking two or three years even for smaller builds.

    Published
    2026-05-12
    Source observed
    , live page
    Access
    full text, public

History

Assessment versions

Recorded events

  1. Assessment publishedBy this site

    First assessment published after both reviews and the final-text check passed.

    About assessment wyea-where-are-all-the-data-centers-5e40a7-a1.