Written by: Long Yue, Wall Street Journal

A company valued at $852 billion, racing towards a trillion-dollar IPO, is being weighed down by its internal chaos.

Since the beginning of this year, OpenAI has undergone nearly five organizational restructures. Executives have been leaving in droves, the security team was disbanded, and employees generally feel exhausted. Meanwhile, the IPO that was originally expected to be completed this year has quietly been postponed to next year.

On August 15, according to the latest report from the Financial Times, multiple insiders indicated that repeated personnel reshuffling has left some employees feeling “frustrated” with this globally most valuable private technology company.

Executives leaving in succession, wave of departures continues

This week, Chief Revenue Officer Denise Dresser announced her departure. She just took office last December, tasked with expanding the enterprise client business, with a tenure of less than a year.

This is the latest case in a series of recent executive departures. As previously disclosed, former Chief Financial Officer and Chief Operating Officer Brad Lightcap, along with ethics chief Chloé Bakalar, have also left, both with a tenure of less than a year.

Previously, Fidji Simo, seen as CEO Sam Altman's "number two" person, transitioned to a part-time consulting role last month after taking medical leave. However, according to two insiders, she still communicates with employees almost every day. One person familiar with the situation stated, "She's still pulling the strings behind the scenes." This individual also described how executives are vying for Altman's attention, with a strong atmosphere of "politics" internally. OpenAI, on the other hand, stated that Simo is fulfilling her responsibilities as a consultant as usual.

A noticeable trend is emerging: several executives have been moved to newly established or vaguely defined positions before formally leaving. Lightcap was shifted to a "special projects" position in April, and Kevin Weil and Josh Achiam also experienced similar transition arrangements before leaving in succession.

Security team disbanded, internal concerns rise

At the end of last month, OpenAI disbanded its "preparedness" team. According to multiple insiders, this team was responsible for assessing whether AI models could pose serious or catastrophic risks and researching response measures. After the disbandment, related responsibilities were distributed among existing teams, with senior employees responsible for different areas like biosafety and cybersecurity.

This was one of the last remaining pieces in OpenAI's original research-led structure. Previously, the "AGI readiness" and "mission alignment" teams had also been disbanded. In 2024, Jan Leike, who co-led the "superalignment" team, resigned and publicly stated that safety issues were being "supplanted by flashy products."

Dylan Scandinaro, head of the preparedness team, has shifted to researching the safety impacts of recursively self-improving AI—systems that can enhance their own capabilities and train other models. Co-founder Greg Brockman stated that OpenAI's technological advancements require "more robust" safeguards, and the company has made adjustments to "integrate research, safety, and security more deeply" into model development.

However, the successive departures of Bakalar, Josh Achiam, and Johannes Heidecke, among others in the security field, have heightened internal unease. One insider candidly remarked, "This is a bit concerning; there is now a sense of urgency to get this right." Another described an internal sense of "a subtle feeling of responsibility and fear, worrying that they are not yet prepared."

In addition, OpenAI recently disclosed a security incident—one model infiltrated another company's system during internal network capability testing, further intensifying external skepticism regarding its security controls.

Revenue growth, but surpassed by Anthropic

On a business level, OpenAI's annualized revenue has increased from $24 billion at the end of last year to about $40 billion this month, marking significant growth. According to insiders, most of the growth came from the new model GPT-5.6 released five weeks ago.

However, competitor Anthropic has been growing even faster. According to data cited by the Financial Times from insiders, Anthropic's annualized revenue surged from $9 billion at the end of 2025 to $47 billion in May of this year, a roughly fivefold increase, and has maintained a similar growth trajectory, surpassing OpenAI.

It should be noted that both companies estimate annual revenue based on recent performance, and there are differences in revenue accounting methods, leading to certain limitations in direct comparisons.

IPO postponed, employees cashing out

OpenAI's IPO was originally expected to be completed this year, but according to insiders, it has now been postponed until next year, when its valuation could reach $1 trillion.

Meanwhile, the company recently completed a nearly $7 billion share buyback, allowing employees to cash out at an $852 billion valuation. According to current and former employees, some people cashed out amounts exceeding $10 million.

However, the costs arising from frequent restructuring are beginning to show. Two insiders indicated that employees generally feel "exhausted" and "frustrated," with some viewing the share buyback as an exit strategy.

Investors have differing views on this. Some investors express concerns about the high turnover rate and the company's lack of focus; others characterize this personnel adjustment as a normal cleanup before the IPO. One investor holding a substantial amount of OpenAI shares stated, "I think it’s not uncommon to reorganize before going public. You need to be clear about who your team is. Some departures are for health reasons, while others are for cleanup."

OpenAI stated that the company is advancing at an "extraordinary speed" and is "not afraid to undergo organizational change."