“Even if these companies were profitable on inference (I do not believe they are), they are dramatically, horrendously unprofitable on subscriptions, and there isn’t a chance in Hell that the majority of those subscriptions convert into token-based API users.”
Our reading:
Anthropic and OpenAI lose a great deal of money on their subscriptions.
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 December 21, 2025 report puts OpenAI's compute margin across paid corporate and consumer products at 70% as of October 2025 [e1]. That measure deducts model-running costs and does not isolate subscription revenue and attributable subscription costs [e1]. It therefore neither establishes nor refutes substantial subscription losses at both companies during October–December 2025 [e1].
Caveats
- The packet lacks matching subscription revenue and cost records for either company for the full target quarter.
- It does not show subscriber usage distributions or allocations of training, free-user costs, and other shared costs.
- The reported compute margin does not establish overall subscription profitability.
Evidence (1 item)
Context only: Fortune / Bloomberg,
the company didn’t release the figures
Why it matters: The article reports improved compute margins across paid corporate and consumer products, while recording that OpenAI did not release the figures. This is relevant countercontext, but neither that broader margin nor the spokesperson's response establishes subscription-specific profitability.
- Publisher
- Fortune / Bloomberg
- 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.
Where we had to interpret
- The companies are Anthropic and OpenAI. The claim concerns each company's subscription business overall, not merely its heaviest users or an individual premium plan.
- The ordinary reading is that subscription revenue falls substantially below attributable costs. No numerical loss threshold or accounting basis is supplied.
- Use October–December 2025, the latest completed calendar quarter before the post that public reporting could describe. This period is an evaluator choice because the statement describes an ongoing state.
- The hypothetical concession about inference profitability is not a condition on the subscription-loss assertion.
Right if: Comparable subscription revenue and attributable cost records show substantial subscription losses at both companies during the target quarter.
Wrong if: Comparable records show that at least one company's overall subscription business was profitable, broke even, or incurred only small losses during the target quarter.
We wrote this test after the outcome was known.
What would settle it
Comparable subscription revenue and attributable cost records for each company for October–December 2025, with a stated accounting basis, subscriber usage distributions, and transparent allocation of shared costs.
We found 38 claims in this post. See them all.
Full record
Source and dates
Where's Your Ed At, Why Are We Still Doing This?, paragraph 50.
- 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 2026-03-17.
Where it was said (1)
Said once in the posts we read.
: Where's Your Ed At, Why Are We Still Doing This?
Even if these companies were profitable on inference (I do not believe they are), they are dramatically, horrendously unprofitable on subscriptions, and there isn’t a chance in Hell that the majority of those subscriptions convert into token-based API users.
- Where
- paragraph 50
- Published
- 2026-03-17
- 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
The final check failed: the write-up said more than the evidence showed.