Greg Brockman walked onto a stage in Napa on Wednesday with a line every founder dreams of delivering: his company, OpenAI, could be ready for an initial public offering in 2027. Asked point-blank whether OpenAI would be IPO-ready next year, the company’s president and co-founder answered, “I think so. Yes.”
It was the headline. It was also the least interesting thing he said all evening.
Because in the same appearance, at a Wall Street Journal tech conference, Brockman admitted his company had spent much of the year losing — and then explained what it had learned from it.
The admission came first
Brockman said OpenAI’s first and second quarters of 2026 were “pretty rough.” The rough patch had a specific shape. In agentic coding — the fast-growing business of AI tools that write and edit software — OpenAI, by its own president’s account, “wasn’t even in the conversation.” The rival setting the pace was Anthropic’s Claude Code, which had gained market share while OpenAI’s own coding agent sat ignored.
“We had Codex, but no one was talking about it,” Brockman said. “We just kept our heads down and executed.”
It is a striking confession from a company that, only a few years ago, defined the entire AI conversation. OpenAI did not merely trail a competitor; its own president said it “wasn’t even in the conversation.” The surveys below put numbers on the gap behind that remark.
The numbers back up the confession
Independent surveys suggest Brockman was not being modest. A First Page Sage survey of 2,514 professional developers, conducted between August and October 2026, put Claude Code at 30.1% of developers’ primary AI coding tools — up 11.6 points since January, and enough to overtake GitHub Copilot as the most-used coding assistant mid-year. OpenAI’s Codex sat at 12.1%.
Stack Overflow’s 2026 developer survey told a similar story from a different angle: 65.5% of respondents who use AI coding agents had used Claude Code in the past year, against 29.5% for Codex. And JetBrains’ Developer Ecosystem Survey found Claude Code in use at work by 39% of professional developers, up from 18% in January, while Codex stood at 16%, up from 3%.
The surveys disagree on absolute numbers — they measure different things — but they agree on direction. Claude Code leads the coding conversation. Codex is growing fast from a low base, nearly tripling its share of developers’ primary tools since January, but it has not caught up.
The lesson: sell trust, not just models
What makes Brockman’s account more than a damage report is the lesson he says the company drew from it. OpenAI’s problem in the enterprise market, as he described it, was not that its models were too weak. It was that big companies did not believe OpenAI was serious about serving them.
“A year ago, we weren’t even really in the enterprise game, right? We were really perceived as a consumer company, and a lot of the companies we’d start talking to would say, ‘Is OpenAI really committed to the enterprise?'” Brockman said.
What changed, in his telling, was OpenAI learning what its enterprise job actually was: to help businesses identify the most important milestones they could achieve with its AI technology, build trust with clients, and demonstrate that it was a reliable partner for the long haul. The comeback, as Brockman tells it, turns on that sales lesson rather than a technology story: the company, he says, learned to prove commitment. Whether that lesson is what drove the growth, rather than new products or other factors, is a question his remarks do not settle.
The third quarter, in Brockman’s telling
Then came the numbers. Brockman said the third quarter brought 70% quarterly growth in annualized revenue, which he attributed to the enterprise side of the business doubling — growth he described as the culmination of investments the company had been making. The majority of OpenAI’s revenue now comes from enterprise customers, he said, where it previously accounted for only around 30% of the total. Consumer revenue, he added, is still growing; the enterprise segment has simply taken off.
The direction of travel is corroborated by OpenAI’s own earlier disclosures. In April 2026, the company said business customers accounted for more than 40% of revenue and were on course to reach parity with consumers by the end of the year. Brockman’s update puts enterprise past the halfway mark ahead of that schedule. He pointed to a recent partnership with JPMorgan Chase as an example of the clients OpenAI is now winning.
A note of caution belongs here, and Brockman’s own remarks supply it. The 70% figure measures growth in an annualized revenue run rate, not revenue actually recognized over the three months, and Brockman gave no dollar total or precise enterprise share. The enterprise financial figures are the company’s own claims, reported secondhand; the developer-survey figures come from independent surveys, and the $50 billion run-rate figure is a Financial Times report rather than an OpenAI disclosure. The arithmetic checks out; the underlying financial performance cannot be independently verified from this package.
What “IPO-ready” does and doesn’t mean
Which brings us back to the headline. “We’ve been very clear this year that we want to go public when we feel is the right time for us, and I feel extremely good about the business,” Brockman said. “I feel like we’ve made so much progress this year, and I’ve honestly never felt more confident in what we’re doing.”
Read carefully, the IPO line is a statement about readiness, not a promise. No listing date has been announced, no exchange chosen, no offering price set. “I think so” is not “we will.” The claim may be doing a different job: it reads as the capstone of the turnaround story he told in Napa — a public marker that, in his telling, the comeback is complete enough to talk about Wall Street. It may equally be routine confidence; his remarks do not settle which.
Whether the comeback is complete is precisely what his remarks leave open. The evidence Brockman offered cannot settle whether the trust lesson caused the third-quarter growth, or whether the coding weakness and the enterprise surge are independent developments that happened to share a year. Enterprise growth could reflect a separate commercial trajectory — cumulative investments and big-client wins — that has nothing to do with Anthropic. And the coding gap itself is unresolved: Codex is gaining, but every survey still has Claude Code ahead.
That is the honest shape of the story Brockman told in Napa. A company that admits it spent the first half of 2026 out of the conversation, says it learned to sell trust instead of just models, and reports numbers consistent with a turnaround — without proving the lesson caused it. The IPO claim is the confident ending he chose. The question now is whether the market, and the developers still choosing Claude Code, buy the ending — or ask to see the next chapter first.
