원문 보기
OpenAI just paid out $7 billion to some of its employees by buying back their shares. The tender offer valued the company at $852 billion, the same as the March round, and OpenAI financed everything itself, instead of bringing in outside buyers. Then, within 72 hours, Brad Lightcap, the COO with 8 years of experience, announced he was leaving.
Two days later, commercial director Denise Dresser also announced her departure after just eight months in the position. Now OpenAI will say that this is a management reshuffle before the IPO. Simply put, getting things in order before what could be the biggest deployment in history. But would key people take the money and walk away if they truly believed the company would soon be worth over a trillion dollars?
No. It looks like there 's trouble in paradise, and Sam Altman can't pretend it isn't. So today we'll look at why some of the people who have the best view of OpenAI's inner workings keep leaving. Why do the official explanations not match and what is such an unstable company doing in preparation for going public?
My name is Guy, and this is the "Financial Bureau". Okay, let's start with this simply colossal $7 billion payout. If you don't know, a tender offer is essentially a way for employees to sell their purchased shares without waiting for an IPO. And OpenAI has conducted them before. The October 2025 tender was about $6.6 billion and was funded by outside investors including Thrive, SoftBank and T.
Rowe Price. But this last case was different. There were no external buyers at all . OpenAI used cash from its March 2026 round—$122 billion raised under the joint leadership of Amazon, SoftBank, and Nvidia—to buy back shares from its own staff. And it did so at the same valuation of $852 billion at which that round closed.
The optimistic interpretation of all this is that the company is being generous with liquidity ahead of the listing. By being self-funded, OpenAI avoided an external valuation that could have changed its capitalization right before the reporting was published. In other words, the cost estimate was not verified.
At the same time, it can be said that a company that seeks to avoid having its value tested is not sure that the next buyer will pay more. So, who actually took the money and left ? Well, there's Kevin Vail, the chief product officer who moved into research to launch "OpenAI for Science," then left in April to found his own AI startup with a valuation of at least $750 million.
Bill Peebles, who ran Sora, and Sinas Naran, CTO of business applications, left around the same time. Fiji Simo joined in May 2025 as General Manager of Applications—a role created specifically for her, when at her peak she was responsible for about two-thirds of the organization. She stepped down from her position in July 2026, after 14 months.
However, I would like to note that she cited a chronic illness as the reason for her departure. So this case is, of course, different. But then we have Chloe Balar, who was hired from Meta in August 2025 as the sole dedicated ethics manager at OpenAI, and in July 2026 she left without any public announcement.
The company's comment on this was that ethics at OpenAI do not belong to just one owner or team. Well, maybe, but it sounds like something they only say after someone has already left. Then, as mentioned earlier, we have Lightcap, the COO with eight years of experience, and Dresser, the CRO who had only been on the job for eight months— both left the same week as Tender.
Now pay attention to what these people have in common, because it's not just about experience. No, these were hired employees, whom Sam Altman touted as evidence of “ adult supervision” and a real management team led by a founder who had already been fired once. Dresser came from a position as head of Slack.
Simo came after leading Instacart, while Balar has been Meta's chief ethicist since 2021, joining their responsible AI team in 2019. And the expiration date of each of these announcements was shorter than a mobile phone contract. So, it seems that OpenAI is having trouble retaining the people it hired to prove that the company has grown up.
It does n't look good. And to better understand the situation, let's take a closer look at the documents regarding each dismissal. Dresser is leaving, and I quote, "to seek other opportunities after the transition period ." Lightcap, meanwhile, wrote that it's a bittersweet moment, but he's moving on to start something new — something that has to do with a few important things the world needs to do right.
And Weil talked about two years expanding consciousness. To be honest , all this is not very convincing unless you are a fan of LinkedIn garbage, because a bunch of top managers cannot spontaneously develop the same thirst for change in such a short period of time. Dresser spent four months consolidating both Lightcap's duties and part of Simo's powers.
She concentrated all this power in her hands, and then she broke free, which is quite telling. Jeff Park of Parafi said, “It’s just unusual to have so many top executive layoffs before a highly anticipated IPO, unless…” And he left the sentence right there . Oh, just a little. Elsewhere, Brian Sullivan said on CNBC that the departure of two top executives in a week at what is arguably the world's most popular company on the eve of an IPO where people are supposed to get rich is, to say the least, surprising to him.
And Kevin McCormack, founder of S Audit, said that if departing executives are not compensated for their losses in their new position, then this is bad news for OpenAI. But with all this, it can be argued that staff turnover is only a symptom, because the work culture and behavior that many now blame for these layoffs were warned about 3 years ago by people closest to Sam Altman.
In November 2023, the nonprofit OpenAI board removed Altman from his position as executive director. The stated reason was not something mundane, such as strategic or commercial differences. The board's accusation was that Altman repeatedly kept them in the dark , undermining their ability to oversee the company and ultimately destroying their trust in him.
He returned 5 days later after 700 of the company's approximately 770 employees signed a letter threatening to switch to Microsoft. Most people dismissed it as a little corporate drama, but the documentation kept coming out. In October 2025, transcripts of testimony released as part of Musk's trial revealed that Ilya Sutzkever, then the chief scientific officer, spent about a year creating a 52- page briefing note on Altman's behavior.
The New Yorker subsequently published a lengthy and thorough investigation, drawing on about 70 pages of internal documents and more than 100 interviews. The first line of the memo described a consistent pattern of lies on Altman's part, undermining the authority of his superiors and pitting them against each other.
The evidence was largely based on screenshots provided by then-CTO Mira Murati. Sutskever is quoted as saying that he did not consider Altman to be a person who should keep his finger on the button. Note that this is an accusation from a battle for control that Altman ultimately won. And the people who nominated them have every reason to try to justify themselves.
But look where they all are now. Sutzkever left in May 2024 and founded Safe Superintelligence. Helen Toner left the board of directors and moved to Georgetown University. Tasha McCauley moved to Rand, while Ian Lake, who co-led the super-alignment direction, stepped down in May 2024, saying security had taken a back seat to shiny products, and he now heads security research at Anthropic.
Murati is gone until September 2024. Every person who officially questioned whether Altman could be trusted to run the company is now gone. But Altman himself, of course, is still in place. Keeping up with everything that's happening in this veritable maze called the "AI market" takes a lot of time and effort , so you should make your life easier by subscribing to analyze the markets and corporate moves that will change your portfolio.
So e QR code on the screen to get started. So, let's go back to OpenAI, which reorganized its security system in July of this year. This meant that the security unit lost its independent reporting and came under the direction of the research director, and almost immediately Johannes Heideka, head of security systems, announced his resignation.
Sandini Agawal, who worked at OpenAI for over six years and previously led the AI safety teams there, also left. Joshua Akayam, the chief futurist who previously led mission alignment, left after nearly 9 years on the job, writing that it was impossible to work on OpenAI's mission outside of the cutting-edge lab.
And at the end of July, the training team, a unit whose main task was to assess catastrophic risks, was disbanded. This is the third security unit to be disbanded since 2024, following a super- agreement in May 2024 and a mission agreement in February 2026. Add to this the history of the company's structure, and the situation looks very strange.
OpenAI was founded on the belief that such a powerful technology was too dangerous to run as a regular company. Then, in October 2025, it lifted the 100x investor return cap, became a Delaware public benefit corporation, and softened the word " safe" in its charter. And then dissolved the team that was supposed to resist into the rest of the company.
And one by one, the fuses that were supposed to make OpenAI special began to disappear. And this is a company that is now asking public markets to trust it and buy shares. The confidential Form S1 was filed with the SEC on June 8, led by Goldman Sachs, Morgan Stanley and JP Morgan. Bankers are talking about a range of $830 billion to $1 trillion, but the timeline has already shifted from the end of 2026 to closer to 2027.
Altman apparently insisted on a trillion and a September window, while CFO Sarah Fryer and the banks lost patience. All of this leads to the main question. Who will really suffer if OpenAI's path to an IPO doesn't hold up? So, over the last quarter, Amazon spent about $54 billion on capital investments , Alphabet—about $45 billion, Microsoft—35 billion, and Meta—30 billion.
That's $165 billion from four companies in 90 days. In the reporting quarter, Nvidia received $81 billion in revenue, of which $75 billion came from data centers. 92% of the revenues of the most valuable company in the world depend on one category of expenses. And a significant portion of that spending is ultimately justified by the bet that companies like OpenAI will be able to grow fast enough to leverage all of that infrastructure.
The Stargate project involves about 7 gigawatts of planned capacity and over $400 billion in capital investment by September 2025. Oracle has a 5-year, $ 300 billion cloud deal starting in 2027. Broadcom, AMD, CoreWeave , Ohio campus with 20- year lease, energy deal in Georgia. All of this is tied to demand from OpenAI.
And at this stage of the AI industry, I probably don't need to point out the cyclical nature of this process. Nvidia invested $30 billion in a March investment round and then sold its chips to OpenAI. Meanwhile, an MIT study found that 95% of corporate pilots in generative AI failed to produce a measurable impact on profits.
So, costs continue to rise , while profits remain purely theoretical, and the entire structure relies on the impeccable work of the company under the leadership of a team that is constantly being fired. The people who held the most information and shares received cash payments based on a valuation set by their own company, and went before outsiders were allowed to look at the financial statements.
Meanwhile, Sam Altman calls this a new chapter for OpenAI, but it seems to be the ninth this year. Eventually, OpenAI will go public, and it will happen at a huge valuation . Whether it will be able to justify this final figure is another question entirely. Well, friends, that's all we have time for today .
What do you think? Is this a company prudently putting things in order before a big IPO, or a company whose executives know they are selling assets at peak value? Let me know what you think about it crazy the hyperscaler capital spending boom is getting , I'll see you in the next episode. Guy was with you, see you later .
Comments 0