Two sets of accounts on the eve of OpenAI's IPO: 400 billion in revenue, 8520 billion in valuation, and a wave of senior executive departures

In August 2026, OpenAI has brought its listing story to the most delicate juncture:

Its annualized revenue has exceeded 400 billion US dollars, with enterprise business surpassing consumer business for the first time, boasting an impressive growth curve. Yet in the same August, the Chief Revenue Officer, long-serving veteran who has been with the company for 8 years, and the second-in-command left one after another, and the safety team also saw a wave of departures. When its $852 billion valuation is about to be tested by the public market, investors are no longer asking "whether AI demand is real", but "whether this company is stable".

OpenAI, the 10-year-old company, is on its closest path to ringing the listing bell in a decade.

In June, it had confidentially submitted S-1 form to the SEC — the listing process has been launched, though the timeline is still undecided. What supports its narrative is a set of hard figures: 40 billion US dollars in revenue, over 20% month-on-month growth in July, and 32% growth rate of enterprise customers. What drags it down, however, is also hard facts: several top executives left within a month, and old governance issues have not been resolved. For OpenAI, the real exam facing its IPO has never been "whether it can go public", but "at what valuation and in what state it will go public".

Revenue: The Truth Behind the Composition of $40 Billion ARR

OpenAI's revenue story hit a key node in mid-August. Bloomberg reports show that OpenAI's annualized revenue run rate (ARR) has exceeded 40 billion US dollars, roughly twice its revenue expectation at the end of 2025. The growth is driven by three forces: the expansion of AI programming software (Codex), subscription sales, and the advertising business that was only tested in ChatGPT in February this year.

The growth rate is equally astonishing. Greg Brockman, co-founder and President, revealed in an internal announcement that the company's annualized revenue in July increased by more than 20% month-on-month; according to investor slides seen by CNBC, enterprise customers grew even faster, reaching 32%.

What is more thought-provoking is the structural change. At the meeting with investors on August 14, CFO Sarah Friar stated that by revenue, enterprise business has now surpassed the ChatGPT-dominated consumer business. "At the beginning of the year, our ratio was 60 (consumers) to 40 (enterprises), but the growth rate of enterprise business far exceeded expectations, and the two have now reversed," she said. This means OpenAI has gone beyond its earlier prediction at the beginning of the year that "the two businesses will be equal by the end of 2026".

Supporting this reversal is the expansion of the scale of enterprise customers: during Denise Dresser's tenure, the enterprise business grew to 2 million customers, doubling over the same period last year; the advertising business is also approaching a 10 billion US dollar run rate.

In other words, OpenAI's 40 billion is no longer just a story of "ChatGPT with 90 million weekly active users", but increasingly resembles the story of an enterprise software company — high profit margin, strong stickiness, and head-to-head competition with Anthropic on the enterprise side.

Valuation and Listing Timing: The $852 Billion Exam Question

Supporting the 40 billion US dollars in revenue is an $852 billion valuation. OpenAI completed its financing in March 2026 with a post-money valuation of $852 billion, making it one of the highest-valued unlisted tech companies in the world. In the secondary market, this price has "remained stagnant" — according to observations from private equity platforms, buyer attention picked up in summer, but quotations still hover around the financing round in March.

The listing itself remains the biggest suspense. OpenAI has confidentially submitted S-1 form to the SEC in June, but has not yet given a specific timetable. At the investor meeting on August 14, when asked about the IPO timetable, executives declined to comment on the grounds that "we have confidentially submitted documents and it is not convenient to discuss"; Brockman also downplayed the threat of Chinese open-source models, saying that "there are misunderstandings about the claim that open source is cheaper". Current information from multiple sources is divergent: The Information once reported that it tends to postpone the IPO to 2027 on the grounds that Sam Altman insists on a "trillion-dollar valuation bottom line"; while Bloomberg and other sources mentioned a timetable within this year. The sharp volatility of SpaceX's stock price after its listing this summer also led the underwriting team to warn that retail investors may lack enthusiasm for OpenAI's issuance.

Talent Drain: The "Huge Warning Sign" on the Eve of IPO

Beyond growth, there is turmoil. This month, OpenAI has seen a wave of intensive executive departures:

• Denise Dresser, Chief Revenue Officer, suddenly announced her departure on August 13, only about 8 months after taking office (she joined from Slack in December 2025). The enterprise business she was in charge of is exactly OpenAI's core business with high profit margin and head-to-head competition with Anthropic.
• Brad Lightcap, a long-serving executive who has worked with Sam Altman for 8 years and former Chief Operating Officer, announced the end of his tenure two days before Dresser left, to "start a new career".
• Fidji Simo, the second-in-command under Sam Altman, took medical leave for health reasons in July and then left the company, remaining as a consultant.

Earlier (in April), four executives including Kevin Weil, Vice President of Science, and Kate Rouch, Head of Marketing, left the company.

The safety and alignment team also saw a wave of departures: Chloé Bakalar, Head of Ethics, Johannes Heidecke, Head of Safety Systems, Joshua Achiam, Chief Futurist, and Sandhini Agarwal, former Head of the AI Safety Team (in July), left one after another.

Kevin McCormick, founder of AI startup SignAudit.AI, said directly on X: "OpenAI's executive departures on the eve of its IPO is a huge warning sign." He pointed out that Dresser left after less than a year in office, and most likely gave up a huge compensation package; "If the departing executives do not get 'full compensation' from their next company, that is bad news for OpenAI."

The turmoil is not an isolated incident. According to two current investors, they were surprised by Dresser's departure, but also said bluntly that "chaos is part of OpenAI's fast-paced culture". Two former employees described that the company has long adhered to the culture of "hiring fast, firing fast", and one of them described it as a "pressure cooker". Earlier departures had their own reasons: Fidji Simo joined OpenAI last year after leaving the position of CEO of Instacart, and left temporarily in July due to "severe deterioration of chronic disease"; among the four executives who left in April, Kate Rouch, Head of Marketing, left to focus on cancer recovery. Wired also reported that the team responsible for the alignment of OpenAI's most powerful models is undergoing chaotic restructuring, and at exactly this time, these models were found to be able to escape the sandbox environment and invade third-party systems — the cracks in governance and security were torn open in the same week.

For Sam Altman and Greg Brockman, this has increased the pressure — investors have already had concerns about competition from Google and Anthropic, the rise of low-cost open-source models, and the stock price volatility of SpaceX in the first two months after its listing.

Strategic Restructuring: Brockman is "Fully Committed"

Amid the turmoil, the power structure is converging towards the founding team. According to Axios, Brockman is "in founder full-immersion mode", contacting customers more frequently than ever, aiming to build a leadership team to surpass Anthropic in enterprise-level applications. He said at the round table in July that his focus last year was on "the computing side and the model execution side", and now he has turned to "computing-driven economy"; another insider regards this round of departures as "a long-overdue cleanup" for underperforming leadership. He publicly explained these departures as a shift in strategic focus: "The way we deploy technology is changing rapidly."

OpenAI appointed Dali Rajic to replace Dresser as Chief Revenue Officer. Rajic was previously President and COO of cybersecurity company Wiz — Wiz was acquired by Google for 32 billion US dollars earlier this year; according to insiders, he was recommended by Josh Kushner, founder of Thrive, an important supporter of OpenAI.

On the product side, OpenAI's strategic narrative is also adjusting. Sarah Friar, Chief Financial Officer of OpenAI, said that "the era of token maximization is over", and enterprise customers have shifted from pursuing token volume to focusing on "the cost per unit of intelligence"; the company emphasized that its latest model has "54% higher efficiency on agent coding tasks", and has cut prices for its model suite. To expand Codex into enterprise workflows, OpenAI also planned to acquire Ona in July this year. On the model side, Codex, which is benchmarked against Claude Code, has reached 5 million users in June — and quickly exceeded 10 million users in late July, and the latest model GPT-5.6 Terra and the budget model GPT-5.6 Luna have also received positive feedback.

Overhanging Governance Shadow

Executive turmoil is not an isolated case, but the latest episode of OpenAI's long-term unstable culture. In 2023, Sam Altman was briefly removed from the position of CEO by the board of directors, and resumed his position a few days later under the pressure of investors and employees; in the high-profile "Musk v. Altman" trial in May this year, this history was mentioned repeatedly — Musk sued OpenAI in 2024 for the company's governance structure, and the jury finally ruled that the lawsuit was filed too late, but Musk vowed to appeal. These old governance issues were brought up again on the eve of the IPO, becoming the background for investors to examine "whether the company is stable".

The details are even more intriguing. The reason for the removal in 2023 was that the board of directors determined that Altman was "not consistently candid" in communication; only a few days later, under the panic of investors and the threat of collective resignation from employees, he regained power. In the "Musk v. Altman" trial in May this year, this past event was repeatedly cited: Musk believed that OpenAI should be a non-profit organization when he invested in it in the early years, but now he accuses it of turning to for-profit and "deceiving" himself. Musk's lawyer also used a memo from Ilya Sutskever, co-founder, to question Brockman, claiming that it pointed out his problems in performance and management — Brockman responded: "I certainly do not agree with that description."

Two Sets of Accounts Before IPO

As of mid-August, OpenAI still has not rung the listing bell. It presents two sets of accounts to the market: one is the growth account of 40 billion US dollars in revenue, enterprise business surpassing consumer business, and +20% growth in July, which proves that AI demand is real and still accelerating; the other is the stability account of $852 billion valuation, successive executive departures, turmoil in the safety team, and unresolved old governance issues, reminding the public market that scale does not equal certainty.

When its revenue has exceeded 40 billion US dollars, the last question OpenAI needs to answer is actually very simple — how many multiples is the market willing to pay for it? And the answer to this question may not be written into the prospectus until the turmoil subsides.

This article is from the WeChat official account "Tech Business", author: Tech Business, published with authorization from 36Kr.