Sam Altman and the company he doesn't own

Hello, and welcome to the very first Founder File. We’re so glad you’re here at the beginning.

At about noon on Friday 17 November 2023, Sam Altman was asked to join a video call. On the other end was most of the board of the company he ran. Ilya Sutskever, OpenAI’s chief scientist and one of its co-founders, told him he was being fired. The board’s stated reason was that he had not been “consistently candid” with them.

Within the hour, the most visible founder in technology had lost the job of his life. Within five days he had it back, the board that fired him was gone, and almost the whole company had signed a letter saying they would follow him out of the door if that’s what it took.

Hold on to that week, because it explains more about Sam Altman than any résumé could. And here is the odd part, the reason he is File #001: the man at the centre of OpenAI, the company behind ChatGPT, has no direct stake in it. His salary is reported at somewhere between $65,000 and $76,000 a year, depending on the source. His fortune, which Forbes put at about $3.3 billion this spring, was built almost entirely elsewhere. And still, when a board tried to remove him, it found that it couldn’t.

This is the story of how that happens: a Missouri kid with a Mac, a first startup that didn’t quite work, a decade running the most famous startup school in the world, and a bet on artificial intelligence so large that it needed more money than any charity could ever raise. (All figures are in US dollars.)

The short version

• Who: Sam Altman, 41, co-founder and CEO of OpenAI.
• What: OpenAI builds the AI models and products behind ChatGPT, which had 900 million weekly users by February this year.
• Why it matters: every other AI founder is now racing, copying or negotiating with the company Altman runs. It is also the clearest case study you will find of leading a company without owning it.
• The number to remember: $0. That is Altman’s direct stake in OpenAI, according to Forbes and the evidence in his own court case.

The founder

Sam was born in Chicago on 22 April 1985, the eldest of four children, and grew up in Clayton, just outside St. Louis. His mother, Connie Gibstine, is a dermatologist, and his father, Jerry, was a real estate broker. When he was eight he was given an Apple Macintosh and taught himself to code, which he has since described as the glory days of computing, when you could make the machine do whatever you wanted and it was just crazy fun. At John Burroughs School he built the school’s website. He came out at 17, after an incident at school around National Coming Out Day, in a part of the world where that was not nothing.

He went to Stanford to study computer science and left after two years, in 2005, to build Loopt, an app that let friends see where each other were. It was a location app before the iPhone existed, and it became part of the very first batch at Y Combinator, the startup accelerator. Loopt raised more than $30 million from investors including Sequoia, never found a mass audience, and was sold to the payments company Green Dot in 2012 for $43.4 million. By Silicon Valley standards it was a modest ending. Reporting since then says that senior colleagues twice asked the board to replace him as CEO, and that the board declined both times.

But here is the thing worth noticing. The real prize from Loopt wasn’t the sale price. It was the people. Paul Graham, Y Combinator’s co-founder, had spotted something early. He compared Altman to a young Bill Gates, and later said you could drop him on an island of cannibals and come back in five years to find him king. Altman became a part-time YC partner in 2011, and in 2014, at 28, Graham handed him the whole thing. His job stopped being to build one company and became to see a thousand of them. He pushed YC towards much bigger batches and “hard technology”, and the alumni list is a roll call of the decade: Stripe, Airbnb, Coinbase, Instacart, Reddit. For eight days in 2014 he was even Reddit’s CEO, a cameo that previewed a habit of stepping into an institution and then keeping the relationship. He stayed on Reddit’s board until 2022.

He left YC in March 2019 to run OpenAI full time. You will sometimes read that he was pushed out. Graham’s own account is that he and Jessica Livingston told Altman that if he was going to work full time on OpenAI, YC should find someone else to run it, and that Altman agreed straight away. They didn’t want him to leave, just to choose.

Away from the work, there is a warmer thread. Altman met his husband, the Australian engineer Oliver Mulherin, in 2015, and they married in Hawaii in January 2024. In May that year they signed the Giving Pledge, promising to give away most of their wealth. Their son was born on 23 February 2025, early, and spent his first days in intensive care. A year in, Altman has said (as Fortune reported) that the careful routine he used to keep, with sleep, weights and meditation, has “all fallen to crap”, that he has accepted a few chaotic years, and that most other things now struggle to beat the bar of time with his family. They spend weekends on a ranch in Napa, hiking where there’s no phone signal. It’s an unexpectedly gentle picture of someone often described as the most driven person in tech.

The problem

By 2015 there was a worry among some of the people closest to AI: that if the technology became as powerful as they thought it might, it would end up controlled by one or two giant companies. Google had bought the lab DeepMind the year before.

In December 2015, Altman, Elon Musk, Greg Brockman, Ilya Sutskever and a small group of researchers launched OpenAI as a nonprofit, with $1 billion in pledges and a mission to build AI for the benefit of everyone. The headline number was bigger than the reality: by 2019, only about $130 million of it had actually been collected.

That gap was the real problem. Training frontier AI models costs money on a scale no charity could match, and a mission that can’t pay for its compute isn’t much of a mission. Musk left the board in 2018. In 2019 OpenAI created a capped-profit arm so it could raise serious capital, Microsoft invested $1 billion, and Altman gave up YC to become CEO. The thing he was really solving wasn’t technical. It was how you fund a moonshot that costs more than almost any startup has ever needed, without losing the thing you set out to protect.

The product

ChatGPT arrived on 30 November 2022 as a modest “research preview”. It wasn’t OpenAI’s first model. It was the first one that ordinary people could simply talk to. A million people signed up in five days. By February this year it had 900 million weekly users, and more recent figures put it close to a billion.

Behind the chat window sits the real product: the models themselves, which OpenAI also sells to developers and businesses through its API, and which Microsoft and others build into their own software. All of it runs on enormous amounts of computing power that OpenAI buys and builds with partners, which is why you hear so much about data centres and a project called Stargate whenever Altman speaks.

His public argument has stayed consistent even when the tactics haven’t: powerful AI is close enough to plan for, the upside is huge, and it is better built in the open than left to a handful of closed labs. Critics hear a sales pitch. Supporters hear the only strategy that matches the cost of the work. Both can be true, because at this price the mission and the fundraising become the same document.

Five days in November

Back to that week in 2023. The board that fired Altman was four people: Sutskever, Adam D’Angelo, Helen Toner and Tasha McCauley. Greg Brockman was removed as chair and promptly quit. Mira Murati, the chief technology officer, became interim CEO. Then the company pushed back. Around 700 of OpenAI’s roughly 770 employees signed a letter threatening to leave. Microsoft’s Satya Nadella announced that Altman would join Microsoft. Sutskever publicly said he regretted taking part.

By the following week Altman was back, with a new board chaired by Bret Taylor. On paper, power at OpenAI sat with a nonprofit board. In practice it sat with the person that the researchers, the product teams and the money would follow. Governance lost to gravity.

Altman has been unusually frank about his own part in it. In April this year he wrote that he was “not proud” of how he handled a conflict with the previous board, and that he is a flawed person who tends to avoid conflict. It is a small admission, and a human one.

The funding

OpenAI’s money has come in stages, each one bigger than the last.

• 2015: $1 billion pledged, around $130 million collected by 2019.
• 2019: Microsoft puts in $1 billion and becomes the company’s key cloud partner.
• October 2025: OpenAI reorganises. The business becomes a public benefit corporation, OpenAI Group PBC, with the nonprofit OpenAI Foundation still in control and holding about 26%, and Microsoft holding about 27%.
• 31 March 2026: OpenAI closes a record $122 billion round at an $852 billion valuation, co-led by SoftBank alongside Andreessen Horowitz, D.E. Shaw, MGX, TPG and T. Rowe Price-advised accounts, with Amazon, Nvidia and Microsoft joining.
• Late September 2026: reports say OpenAI is in early talks to raise at least $30 billion more at a valuation of around $1.4 trillion. Altman has said an IPO this year would be “ill-advised”, given the current mood around AI safety, and that it won’t happen in 2026.

Notice who is missing from that cap table. Altman is. Court filings and Forbes show no direct OpenAI equity, and he has said as much. His own money sits in a private portfolio of more than 400 companies. The biggest disclosed position is Helion Energy, a fusion-power startup he has backed for over a decade, put at around $1.7 billion in filings from the Musk trial (as of the end of 2025). Others include early stakes in Stripe and Reddit and the longevity company Retro Biosciences, where he put in $180 million of his own money.

The business model

OpenAI earns money in four ways: paid ChatGPT subscriptions for individuals, subscriptions and contracts for businesses, access to its models through the API, and, new this year, advertising.

The ads are the story of 2026. They began testing in the US in February, appear at the bottom of ChatGPT’s answers, are clearly labelled, and are not shown to under-18s. OpenAI says they don’t influence the answers. By late August the ad business had reached a $1 billion annualised run rate in about 200 days.

The overall numbers are huge and so are the costs. OpenAI made $13.1 billion in revenue in 2025 and burned about $8 billion. Its annualised revenue is now reported to be nearing $70 billion. It forecasts more than $280 billion of revenue in 2030, split roughly evenly between consumers and businesses. But it also expects to spend around $600 billion on computing by 2030, down from the $1.4 trillion Altman talked about last November, and the Financial Times reports its own projections show a cumulative cash burn of about $278 billion between 2026 and 2030.

That is the business model in one sentence: spend almost unimaginable sums now, on the bet that revenue arrives faster than the bills.

Altman’s own business model is different, and quietly clever. He didn’t make his money by owning the thing everyone uses. He made it by investing in the things that thing needs. Compute needs energy, so there is Helion, and earlier the nuclear company Oklo, which he chaired until April 2025. A web full of convincing machines needs a way to prove you’re human, so in 2019 he co-founded Tools for Humanity, whose World project scans irises to issue a proof-of-personhood credential. Regulators in several countries have objected to that rollout, and it hasn’t yet settled whether it’s infrastructure for the future or a solution looking for permission.

Which brings up the obvious awkward question. In March this year Altman stepped down as Helion’s board chair as the fusion company entered early talks to supply OpenAI with power (reportedly 12.5% of its output, five gigawatts by 2030). He still holds the financial stake, but says he has recused himself from the negotiations. He made the same move at Oklo. Both are sensible. Neither makes the overlap go away.

What could go wrong

Trust. In April, the New Yorker published a 16,000-word investigation by Ronan Farrow and Andrew Marantz, based on more than a hundred interviews, asking whether Altman can be trusted with so much influence. Many of those sources raised concerns about candour. Altman answered in a blog post, accepting that he handles conflict badly. The question is not settled, and it follows him into every negotiation.

The money. A company burning this fast depends on investors continuing to say yes. The $600 billion reset shows that even OpenAI is adjusting its ambitions to the market. If revenue growth slows, the bills don’t.

The founding story. Musk sued, seeking up to $150 billion and arguing that OpenAI betrayed its nonprofit mission. In May 2026 a jury ruled unanimously, in under two hours, that he had waited too long to sue. That was a win on timing rather than on the underlying question, and Musk has said he will appeal.

Safety, and the pace of it all. In mid-September, Anthropic’s Dario Amodei published an essay urging AI labs to slow the pace of their most advanced models. Both Altman and Musk backed it, and Altman said OpenAI would give independent evaluators employee-like access to its work, with details to follow. It was a notable change of tone from a company known for shipping fast. Caution as policy, acceleration as strategy: that is the Altman tension in one sentence.

The human cost of being him. On 10 April a man threw a Molotov cocktail at Altman’s San Francisco home, setting fire to part of a gate. No one was hurt. The suspect, a 20-year-old, had reportedly written essays about AI ending humanity. Altman responded by posting a photo of his family and asking everyone to turn down the rhetoric. A second incident followed days later. Whatever you think of him, the stakes of this story have become very personal.

What to watch

1. The next round. Does the $30 billion-plus raise at around $1.4 trillion close, and on what terms? It is reportedly meant to be the last before an IPO. Altman has said “not 2026”; 2027 is the widely reported target, though nothing is confirmed.
2. Revenue against burn. Does the run rate near $70 billion keep climbing, and do advertising and business customers grow fast enough to bend the loss curve before the $278 billion projection comes true?
3. The evaluators. Altman has promised independent access “soon”. Details, who gets in, and what they can actually see will show whether the September slowdown is a policy or just a mood.
4. The conflicts. Watch the Helion power deal, and every other tie between OpenAI and the companies Altman owns a piece of. How he handles those will tell you a lot about the governance he built.
5. Musk’s appeal. It probably won’t change OpenAI’s trajectory in the next year, but it keeps the founding bargain in the news.

---

That’s File #001. Thank you for reading it. If it made you think of someone, forward it along, and if there’s a founder you’d like to see opened next, tell us at info@aifounderfiles.com. File #002 lands next Sunday.

Facts in this file are drawn from the sources below. Where a figure is reported rather than confirmed by the company, we say so. Spotted a mistake? Email info@aifounderfiles.com and we’ll correct it.