“This means that OpenAI would likely be burning more like $6 billion a year on server costs if it wasn’t so deeply wedded to Microsoft — and that's before you get into costs like staffing ($1.5 billion a year), and, as I've discussed, training costs that are currently $3 billion for the year and will almost certainly increase.”
Our reading:
OpenAI's training costs, currently $3 billion a year, will almost certainly increase; without Microsoft's discount its server costs would likely be about $6 billion a year.
Assessment: Unresolved
What happened
Did the forecast come true?
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
Only the training-increase forecast is scored. The accessible annual figures show higher R&D compute spending, including $8.3 billion in 2025, but do not clearly reconcile that category with the $3 billion training baseline. The original report that may establish the comparison could not be read. These gaps prevent a decisive finding; they are not evidence that training costs failed to rise.
Caveats
- The distinction between training and broader research spending materially affects the comparison.
- Experimental training counts; this assessment does not restrict training to final runs.
- The qualified $6 billion server-cost counterfactual is not scored.
Evidence (1 item)
Context only: Stanford Institute for Human-Centered Artificial Intelligence,
OpenAI’s reported compute spend
Why it matters: The figure reports $8.3 billion of OpenAI R&D compute spending in 2025. This is consistent with rising training expenditure, but the chart's broader category does not by itself establish a like-for-like increase from the $3 billion training baseline.
- Publisher
- Stanford Institute for Human-Centered Artificial Intelligence
- Kind of source
- Secondary summary
- Measures
- OpenAI, Annual R&D compute expenses, USD billions per calendar year, estimate, 2025-01-01 to 2025-12-31
- About events on
- Published
Probably right?
One AI thinks this is probably right, and the other says it’s too early to tell. The evidence isn’t conclusive, so the verdict above stays unresolved. It counts only when you pick Speculative under “How sure?”.
- Anthropic’s AI: The 2026 Stanford AI Index reports OpenAI's research compute at $8.3B in 2025 and says its compute spend rose sharply from 2024. The category is broader than training, but the amount is far above the $3B baseline.
How we judged it
- Kind of claim
- Forecast
- Deadline
- None given. We don’t add one.
Where we had to interpret
- Score only the training-cost increase forecast. The qualified server-cost counterfactual is outside the scored commitment.
- The baseline is the cited $3 billion annual training-cost estimate for 2024.
- An increase means a subsequent calendar year's aggregate training costs exceed that baseline on a comparable basis.
- Training includes experimental and unsuccessful training work; it is not restricted to final runs for released models.
- No deadline is supplied. A later spending projection alone does not establish that the increase happened.
- A broader R&D total requires a reconciliation to the baseline training category before it can settle the forecast.
Right if: Reliable reporting establishes that OpenAI incurred more than $3 billion in comparable annual training costs in a calendar year after 2024.
Wrong if: A definitive end to OpenAI's model-training activity, with a complete comparable spending record, establishes that no subsequent annual training total exceeded $3 billion. A temporary decline or lack of an increase so far is insufficient.
We wrote this test after the outcome was known.
What would settle it
Readable company records or credible dated reporting showing actual post-2024 annual training costs above $3 billion, with sufficient category detail to compare them with the baseline.
We found 56 claims in this post. See them all.
Full record
Source and dates
Where's Your Ed At, The Subprime AI Crisis, paragraph 22.
- 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 2024-09-16.
Where it was said (1)
Said once in the posts we read.
: Where's Your Ed At, The Subprime AI Crisis
This means that OpenAI would likely be burning more like $6 billion a year on server costs if it wasn’t so deeply wedded to Microsoft — and that's before you get into costs like staffing ($1.5 billion a year), and, as I've discussed, training costs that are currently $3 billion for the year and will almost certainly increase.
- Where
- paragraph 22
- Published
- 2024-09-16
- 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.