“Even if revenues were to match those that had been reported, OpenAI’s inference spend on Azure consumes them, and appears to scale linearly above revenue.”
Our reading:
Even at reported revenue levels, OpenAI's Azure inference spend consumes its revenue and appears to scale linearly above it.
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
Epoch AI presents about $1.8 billion in estimated 2024 inference spending, but cautions that its figures partly rely on projections for the final quarter [e1]. Ars Technica reported on June 16, 2026 that 2024 financial statements showed $2.65 billion in cost of revenue and $3.7 billion in revenue; the report does not establish total Azure inference spending [e3]. Entrepreneur reported $4.3 billion in revenue for January–June 2025 without a corresponding Azure inference spending figure [e2]. The lower 2024 figures weigh against the claim, but neither measures the required realized Azure inference total, and the H1 2025 comparison also remains unresolved [e1, e2, e3].
Caveats
- The 2024 inference estimate partly uses projections and does not establish realized Azure inference spending [e1].
- Reported cost of revenue is not reconciled with Azure inference spending, including credits, payment timing and serving costs potentially classified elsewhere [e3].
- The H1 2025 revenue report provides no matching Azure inference spending total [e2].
Evidence (3 items)
Context only: Epoch AI,
They should not be considered completely reliable; for one, they rely on projections for the final quarter of 2024.
Why it matters: Epoch summarizes a roughly $1.8 billion inference estimate for 2024, materially below the claim's revenue benchmark. Its explicit reliance on projected spending prevents treating it as a settled measurement of realized Azure expense.
- Publisher
- Epoch AI
- Kind of source
- Secondary summary
- Measures
- OpenAI, Estimated inference compute spending, partly projected, USD billions, estimate, 2024-01-01 to 2024-12-31
- About events on
- Published
Context only: Entrepreneur,
OpenAI’s revenue for the first half of 2025 was 16% more than its revenue for all of 2024.
Why it matters: The article reports the $4.3 billion H1 revenue comparator, attributing it to shareholder documents viewed by The Information. It supplies no corresponding Azure inference expense, so it cannot settle the comparison.
- Publisher
- Entrepreneur
- Kind of source
- News report
- Measures
- OpenAI, Revenue, USD billions, estimate, 2025-01-01 to 2025-06-30
- About events on
- Published
Context only: Ars Technica,
But the company’s fast-growing revenues are still dwarfed by its even more significant expenses.
Why it matters: The article reports audited 2024 cost of revenue of $2.65 billion against revenue of $3.7 billion. This is material contrary context, but it does not establish that all Azure inference spending is included in cost of revenue.
- Publisher
- Ars Technica
- Kind of source
- News report, audited
- Measures
- OpenAI, Cost of revenue, USD billions, company reported, 2024-01-01 to 2024-12-31
- About events on
- Published
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Conditions
- even if reported revenues are accurate (their words)
Where we had to interpret
- Only the firm 'consumes them' clause is graded. Its ordinary meaning here is that Azure inference spending equals or exceeds the reported revenue being compared.
- The expressly hedged linear-scaling clause is excluded from binary grading.
- The comparison uses CY2024 and H1 CY2025 separately: the context supplies reported revenue comparators for those periods. September spending cannot be compared with six-month revenue.
- 'Even if' grants the reported revenue amounts for the comparison; it is not a forecast whose antecedent must occur.
- The contract concerns reported estimates of realized spending and revenue, not annualized run rates or total compute commitments.
Right if: Independent, comparable evidence establishes that Azure inference spending was at least $3.7 billion in CY2024 and at least $4.3 billion in H1 CY2025.
Wrong if: Independent, comparable evidence establishes Azure inference spending below the corresponding reported revenue benchmark. Opposite findings for the two periods would produce a mixed assessment.
We wrote this test after the outcome was known.
What would settle it
Authenticated records or independently checkable financial schedules establishing realized Azure inference spending separately for CY2024 and H1 CY2025, excluding training, research, other providers and commitments, and reconciling credits, cash payments, expense recognition and free-user serving costs. Those totals would be compared with the respective $3.7 billion and $4.3 billion revenue benchmarks.
We found 14 claims in this post. See them all.
Full record
Source and dates
Where's Your Ed At, Exclusive: Here's How Much OpenAI Spends On Inference and Its Revenue Share With Microsoft, paragraph 88.
- 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 2025-11-12.
Where it was said (1)
Said once in the posts we read.
: Where's Your Ed At, Exclusive: Here's How Much OpenAI Spends On Inference and Its Revenue Share With Microsoft
Even if revenues were to match those that had been reported, OpenAI’s inference spend on Azure consumes them, and appears to scale linearly above revenue.
- Where
- paragraph 88
- Published
- 2025-11-12
- 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.
Held backBy us
A reviewer said our test wasn’t fair to what was actually said.