OpenAI president Greg Brockman said the company's first half of 2026 was "pretty rough," then claimed the third quarter brought "70% quarterly growth in our annualized revenue," with enterprise "doubling." He spoke at the Wall Street Journal's WSJ Tech conference on November 4, 2026. The words are on the record. The numbers behind them are not: no dollar base, no audit, no filing.
This post separates what he said from what is verified, and shows how the claim sits next to the other revenue numbers we have tracked, including the $50 billion versus $70 billion run-rate dispute and Anthropic's pre-IPO numbers.
TL;DR: claimed versus verified
| Claim | Who said it | Status |
|---|---|---|
| Q1 and Q2 2026 were "pretty rough" | Brockman, WSJ Tech clip | On the record, quoted from WSJ's caption file |
| In coding, "we weren't even in the conversation" | Brockman | On the record |
| 70% quarterly growth in annualized revenue in Q3 | Brockman | Spoken, unaudited, no dollar base |
| Enterprise side "doubling" | Brockman | Spoken, unaudited |
| Majority of revenue now from enterprise | Reuters report of the session | Reported, no percentage given |
| Ready for an IPO in 2027 | Brockman: "I think so" | Statement of readiness only |
| GPT-5.6 Sol and ChatGPT Work drove the turn | Brockman named them | His attribution, not independently measured |
What did Brockman actually say?
The most careful account we found is by FourWeekMBA. It says it worked from the Wall Street Journal's own English caption file for a 71-second clip, published at 02:04 UTC on November 5, and that it did not read the full session. The clip is titled "Greg Brockman on the Rise of Anthropic." The interviewer asked when he realized Anthropic would overtake OpenAI in revenue.
According to that caption file, Brockman answered that it "became fairly clear, honestly, even at the very beginning of the year." He then described the first half: "Q1 was pretty rough. Q2 was pretty rough. It really looked in many ways like coding, we weren't even in the conversation. We had Codex, but no one was talking about it."
On the third quarter he said: "Q3, entirely different. We had 70% quarterly growth in our annualized revenue. And if you just break it down by the enterprise side, doubling." He tied the turn to earlier investment: it "wasn't something we did in that quarter," and he named "a Frontier model out in the form of GPT-5.6 Sol" and "ChatGPT work released as well."
We could not open the Wall Street Journal article itself, so quotes here come from the secondary account of the caption file. Treat the wording as relayed.
What does "70% growth in annualized revenue" mean?
Annualized revenue, also called run rate, takes a recent period and multiplies it up to a year. If a month brings in $4 billion, the run rate is $48 billion. It is a projection, not booked sales. Two things follow.
First, "70% quarterly growth" most plausibly means the run rate at the end of Q3 was about 1.7 times the run rate at the end of Q2. Brockman did not give the base, so no one outside OpenAI can compute a dollar figure from his sentence.
Second, the definition of revenue matters. We covered how the same company's September figure was reported as about $50 billion in one count and about $70 billion in another, with sources saying the higher number included partner gross revenue. If you want to compare labs, you must know whether each number is gross or net. Brockman's sentence does not say.
If enterprise annualized revenue doubled while the total grew 70%, the non-enterprise remainder must have grown by less than 70%. FourWeekMBA draws the same inference. It is arithmetic on spoken numbers, so it is a rough check and not a result.
How does this compare with Anthropic?
FourWeekMBA sets Brockman's number next to Anthropic's amended S-1, filed on November 2. It reports quarterly revenue of $4,731 million, $11,555 million and $16,956 million for the quarters ending March 31, June 30 and September 30, 2026. By its arithmetic, that is growth of about 144% in Q2 and about 47% in Q3.
Do not read that as "OpenAI grew faster." One figure is reported quarterly revenue, unaudited, in a registration statement amendment. The other is a spoken growth rate in run-rate terms, with no dollars. The gap of 23 percentage points is the outlet's arithmetic across two different measures, and the outlet says so. We have not independently opened the S-1/A for this post.
What the two sources do agree on is the direction of the contest. Brockman's own interview framing is that Anthropic overtook OpenAI in revenue this year. Our coverage of the filing's context is in the Anthropic IPO valuation reports.
What does the enterprise shift look like over time?
| Date | What was reported | Source type |
|---|---|---|
| March 2026 | OpenAI said enterprise was over 40% of revenue and on track for parity with consumer by year end | Company statement, as reported by outlets |
| Aug 13, 2026 | Axios reported OpenAI expects enterprise to make up about half of revenue by year end, and that Brockman took on more operating responsibility | Axios |
| Mid-Aug 2026 | CNBC reported CFO Sarah Friar told investors "those lines have now crossed" | Cited by FourWeekMBA |
| Oct 8, 2026 | Reports of about $50 billion annualized at end of September, not $70 billion | Reuters, CNBC, FT |
| Nov 4, 2026 | Brockman: majority from enterprise, Q3 annualized growth 70%, enterprise doubling | WSJ Tech session |
The timeline shows one consistent claim: enterprise passed consumer sometime this summer. The new part is the Q3 growth rate and Brockman's account of why. The enterprise crossover itself was already reported in August.
What drove the turn, in Brockman's account?
He named two launches. GPT-5.6 Sol is the frontier model OpenAI released in August. ChatGPT Work is the workplace agent product that handles longer tasks across threads. He also said coding was the weak point in the first half, which fits earlier reporting that OpenAI missed monthly revenue targets after losing ground to Anthropic in coding and enterprise.
His line "We had Codex, but no one was talking about it" is a frank admission, and it is consistent with the Cat Wu hire the same week. Business Insider reported that Wu will lead product management for OpenAI's core products, and framed the move as a possible boost for Codex adoption. That framing is the outlet's view.
One caution: Brockman's attribution to specific launches is his own. Revenue growth in a quarter has many causes, including sales hires, pricing, bundling with partners and plain demand for agents. A founder on stage is a primary source for what he believes, not for why customers bought.
What is not known
- Dollar figures. No revenue amount for Q3 or for the end of the quarter was given in the sources we read.
- The definition. Gross or net, and whether partner revenue is counted, is not stated.
- The enterprise share. Reuters, as relayed, says "majority" with no percentage.
- The IPO date. "I think so" about readiness in 2027 is not a plan. Other reports we saw conflict on timing, so we leave it unsettled.
- Independent verification. We found no filing or audited statement from OpenAI that backs the 70%.
What this means for people who build or pay
This is a finance-adjacent story, so here is the practical consequence. A lab that says its enterprise side doubled in a quarter is investing in agents for work, coding and long tasks. For builders, that points to continued investment in Codex and ChatGPT Work, and likely more enterprise-style features such as admin controls, usage reporting and bundled seats.
For pricing, nothing in the sources announces a change. But when two labs compete for the same business customers and both are close to going public, watch for changes in usage limits, seat prices and credits. Our coverage of the Brockman PHP2Shell remarks and the earlier hedonic treadmill post shows how he uses public posts to frame OpenAI's direction. This session follows that pattern.
How to check claims like this yourself
- Ask for the base. A percentage with no dollar amount cannot be compared. Wait for a filing or a named figure.
- Check gross versus net. The same company produced $50 billion and $70 billion for one month. Definitions decide the number.
- Match the measure. Run-rate growth and reported quarterly revenue are not the same thing. Compare like with like.
- Find the primary text. The WSJ clip and the full session are the primary sources here. Quotes in outlets are only as good as their transcripts.
- Look for the filing. Anthropic's S-1/A gives dollar figures (unaudited, per FourWeekMBA). OpenAI has not published equivalent numbers in anything we read.
Bottom line
Brockman said something notable: a rough first half, then a sharp third quarter led by enterprise, credited to GPT-5.6 Sol and ChatGPT Work. The candor is real and the quotes are traceable to a WSJ caption file. The growth numbers are spoken, unaudited and without a base. Treat them as a signal of direction, not as a revenue figure, until OpenAI or a filing supplies the dollars.
Related reading
- OpenAI says $50B run rate, not $70B: claimed vs verified
- Cat Wu leaves Anthropic and joins OpenAI
- Anthropic IPO valuation reports: claimed vs verified
- GPT-5.6 Sol and Luna in ChatGPT
- ChatGPT Work: thread orchestration and mobile
- Brockman on the hedonic treadmill of model expectations
- Sources: FourWeekMBA on the WSJ Tech clip, Axios on OpenAI's revenue leadership
Details are accurate as of November 5, 2026. We will update this post if OpenAI publishes figures or the full session transcript.