“To be clear, the point I’m making is not that there’s no demand, just that the vast majority of that demand is coming from either Anthropic and OpenAI — two companies that cannot afford to pay for it long-term — and hyperscalers, who are mostly buying compute on behalf of OpenAI and Anthropic.”
Our reading:
There is some AI compute demand, but the vast majority of it comes from Anthropic and OpenAI, two companies that cannot afford to pay for it long-term, and from hyperscalers, who are mostly buying compute on behalf of OpenAI and Anthropic.
Assessment: Unresolved
What happened
Was it true for the period described?
Outcome: Unresolved
We checked, and the evidence so far doesn’t settle it.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
The claim says the vast majority of AI compute demand comes from OpenAI and Anthropic, or from hyperscalers mostly buying on their behalf. Testing it needs shares of compute demand by end buyer for 2026-01-01 to 2026-07-22, and the end use of capacity that hyperscalers rent. No source in the packet gives either. On 2026-04-09 Meta committed another $21 billion to CoreWeave for dedicated AI cloud capacity running from 2027 through December 2032 (e1). On 2026-04-15 Jane Street committed about $6 billion to CoreWeave's cloud services (e2). These are large purchases by buyers other than OpenAI and Anthropic. They do not show what share of total demand those buyers represent. The report does not say whether Meta's capacity serves Meta's own products or other customers (e1). Purchases by other buyers can coexist with the claimed concentration. The claim that OpenAI and Anthropic cannot afford to pay long-term is a forecast with no stated horizon, so it is not scored here.
Caveats
- The packet covers two deals, both with CoreWeave, and gives no market-wide total to compare them against.
- The end use of Meta's CoreWeave capacity is not stated.
- Neither deal shows how much was spent within the target period. Meta's added capacity runs from 2027.
- No evidence shows what share of hyperscaler-rented capacity serves OpenAI and Anthropic.
Evidence (2 items)
Context only: The Next Web,
Meta committed an additional $21 billion to CoreWeave for dedicated AI cloud capacity running from 2027 through December 2032
Why it matters: Meta is a very large compute buyer. The report does not say whether this capacity serves Meta itself or others, so it does not settle the 'on behalf of OpenAI and Anthropic' part.
- Publisher
- The Next Web
- Kind of source
- News report
- About events on
- Published
Context only: Reuters via The Star,
Jane Street also made a $1 billion equity investment in CoreWeave
Why it matters: Shows some demand from outside the named group. One deal says nothing about the overall shares of demand.
- Publisher
- Reuters via The Star
- Kind of source
- News report
- About events on
- Published
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Conditions
- not that there's no demand (their words)
- hyperscalers mostly buying on behalf of OpenAI and Anthropic (their words)
Where we had to interpret
- Testable core: in 2026 up to the post, the vast majority (read as well over half) of AI compute rental demand came from OpenAI and Anthropic directly, or from hyperscalers buying mostly to serve OpenAI and Anthropic.
- 'Cannot afford to pay for it long-term' is a forward-looking judgment about solvency. It is noted but not scored here, because no period or test is given.
- Target period: 2026-01-01 to 2026-07-22, the most recent state before the post.
Right if: Spending or contract data show OpenAI, Anthropic, and hyperscaler purchases made mainly for them account for the large majority of AI compute demand.
Wrong if: Data show a large share of AI compute demand comes from other buyers, or that hyperscalers mainly buy compute for their own use (for example Meta, or Microsoft's own products) rather than for OpenAI and Anthropic.
We wrote this test after the outcome was known.
What would settle it
Credible market-wide data on AI compute spending or contracted capacity by end buyer for the first half of 2026, plus disclosure of what hyperscaler-rented capacity was used for.
We found 37 claims in this post. See them all.
Full record
Source and dates
Where's Your Ed At, The Subprime Data Center Crisis, paragraph 63.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2026-07-22.
Where it was said (1)
Said once in the posts we read.
: Where's Your Ed At, The Subprime Data Center Crisis
To be clear, the point I’m making is not that there’s no demand, just that the vast majority of that demand is coming from either Anthropic and OpenAI — two companies that cannot afford to pay for it long-term — and hyperscalers, who are mostly buying compute on behalf of OpenAI and Anthropic.
- Where
- paragraph 63
- Published
- 2026-07-22
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Unresolved. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.
Held backBy us
A reviewer said our test wasn’t fair to what was actually said.