“It doesn’t just cost more to run OpenAI than it makes — it costs the company a billion dollars more than the entirety of its revenue to run the software it sells before any other costs.”
Our reading, the sentence that was checked:
In 2024 OpenAI's compute costs alone exceeded its entire revenue by about $1 billion.
Assessment: Unresolved Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, There Is No AI Revolution, paragraph 20.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-wheres-the-money-0fed67-a1
We recorded 42 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 2025-02-24. 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
- a billion dollars more than the entirety of its revenue
- before any other costs
- $3 billion
- $2 billion
- 2024
Our interpretation
- The surrounding passage explicitly adds model-training and model-running compute costs. Assess that combined amount, rather than inference alone.
- Compare those costs with total calendar-2024 revenue, not subscription revenue or annualized revenue.
- Read the $1 billion difference as approximate.
- Separately identified research compute must be reconciled with training compute before adding it; it must not be silently omitted or double-counted.
- Require completed-year estimates or results, rather than treating forecasts as realized spending.
Counts as supported if: Comparable completed-year figures show that 2024 training and inference compute costs exceeded total revenue by approximately $1 billion.
Counts as contradicted if: Comparable completed-year figures show no excess, or an excess materially different from approximately $1 billion.
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 cited estimates reproduce the claim's arithmetic: $3 billion of training plus $1.8 billion of inference, less $3.7 billion of revenue, gives about $1.1 billion. But Epoch says the spending figures relied on projections for the final quarter. It also identifies separately amortized research compute. The evidence explains the claim but does not settle the completed-year difference on a consistent cost definition.
Limitations
- The available breakdown ultimately relies on press accounts of investor documents.
- Broader research-and-development expenses cannot be substituted for compute expenses.
Evidence (2 items)
Context only: Epoch AI — Josh You,
$3 billion on training compute, $1.8 billion on inference compute
Why it matters here: The training and inference components total $4.8 billion. The separate research category makes the definition of total compute spending material, and e2 explains why these figures are not completed-year confirmation.
- Publisher
- Epoch AI — Josh You
- Kind of source
- Secondary summary
- Measures
- OpenAI, Estimated training, inference, and amortized research compute expenses, USD billions, estimate, 2024-01-01 to 2024-12-31
- Arithmetic
- 3.0 + 1.8 = 4.8 billion (inputs: Training compute: $3 billion; Inference compute: $1.8 billion; No additional rounding.)
- Describes events of
- Public from
Context only: Epoch AI — Josh You,
They should not be considered completely reliable; for one, they rely on projections for the final quarter of 2024.
Why it matters here: The reported revenue produces a $1.1 billion gap when compared with e1's training and inference estimates. Epoch explicitly warns that the spending figures include projections, so that calculation does not establish the realized gap.
- Publisher
- Epoch AI — Josh You
- Kind of source
- Secondary summary
- Measures
- OpenAI, Reported annual revenue compared with partly projected compute expenses, USD billions, estimate, 2024-01-01 to 2024-12-31
- Arithmetic
- 4.8 - 3.7 = 1.1 billion (inputs: Training and inference estimate from e1: $4.8 billion; Reported 2024 revenue: approximately $3.7 billion; $1.1 billion rounds to $1 billion to the nearest billion.)
- Describes events of
- Public from
What would settle it
A completed-year breakdown of training, inference, and research compute costs, reconciled to total 2024 revenue and the treatment of amortization and cloud credits.
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, There Is No AI Revolution
It doesn’t just cost more to run OpenAI than it makes — it costs the company a billion dollars more than the entirety of its revenue to run the software it sells before any other costs.
- Where
- paragraph 20
- Published
- 2025-02-24
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-wheres-the-money-0fed67-a1: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.