Musk vs
OpenAI,
the true story.
Monday morning, April 27, 2026. Oakland, California. Nine people picked at random sit down in a courtroom. In front of them, on the right, the richest man in the world, glaring. On the left, his former best friend, now his enemy. In the center, the judge. And between the two, ten years of a friendship that started over a dinner and ends over one hundred thirty-four billion dollars. There it is. This is the story I'm going to tell you. Like a novel, because that's what it is.
What you'll walk away with
- Two men, a friendship, a promise: in 2015, they create a lab meant to save humanity from artificial intelligence. Not a penny for shareholders, just for the common good.
- Three years later, one wants to seize power, the other says no. The friend storms off, slamming the door. Seven years pass. The nonprofit turns into a billion-dollar machine. ChatGPT changes the world. Microsoft joins the dance. The friend who left watches, stews, and prepares his revenge.
- This week, in Oakland, they face each other again. Him as the plaintiff. The other on the witness stand. One hundred thirty-four billion on the line, the question of who controls the next decade, and a verdict expected in late May. This is the trial I'm going to tell you about.
Before you sit down
I'm not a lawyer. Not a journalist. I'm just someone who, like you maybe, has been seeing these headlines go by for two years: "Musk sues OpenAI," "Altman turns down a 97 billion offer," "the trial of the century kicks off Monday." And every time, I thought: "But who are these people, and why are they fighting?"
So last night, I made myself a coffee and asked Claude: "Tell me this story as if you were telling it to me at a bar. From the beginning. No technical stuff. With the real people, the real moments, the real blowups." Four sub-agents dug in parallel for two days. When I reread all of it, I had goosebumps.
Because it's not a business affair. It's a story of friendship, of a promise, of betrayal, of payback. A story where the two men saw each other as brothers, and where one today calls the other "Scam Altman" on Twitter while the trial is underway. The key point is that this story is going to decide who controls artificial intelligence over the next ten years. You, your job, your kid. So you might as well know it.
I'm going to tell it to you like a novel. With the scenes, the real dialogue (the emails were made public in 2024 by OpenAI, they're out there on the web), the characters, the moments where it all tips over. If you're an expert on the case and you spot a blunder, reply to my email. I read everything, I fix it.
This story extends the ones I wrote about my son's world 20 years from now with AI and about China's 2026-2030 plan. Three angles on the same big question: who's going to hold the wheel of artificial intelligence?
My bet in 15 seconds
On the courtroom side, I think Musk is going to lose. The 2017 emails that came out are crushing: he wanted himself to turn OpenAI into a commercial company back then, and he wrote it in black and white. On the moral side, he still has a point: OpenAI did betray what it promised at the start. My bet: a splitting-the-difference verdict, symbolic damages, and the definitive end of the "we do AI for humanity" model. From May on, artificial intelligence will be about the money. Period.
What you won't find here
I'm not going to tell you in detail about the failed coup of November 2023 (the five days when Altman got fired and then was brought back in triumph). I'm also not going to talk about the tech behind ChatGPT, or American politics under Trump. Three other stories, all just as juicy. I'll come back to them.
The recap · to get your bearings
Before I dive into the thick of it, here's the arc of the story in seven acts. You get your bearings, and then I'll walk you through each act in detail.
2015 · The alliance. Sam Altman (who runs the biggest startup incubator in Silicon Valley) and Elon Musk reach an agreement over a dinner. They create a company called OpenAI. Not a normal company: a nonprofit, like the Red Cross. They don't want to make money. They just want to stop Google from becoming all-powerful thanks to artificial intelligence. Musk puts in close to 45 million out of his own pocket.
2018 · The break. Three years later, Musk wants to run it alone. He proposes to his partners: "Give me the keys, or fold OpenAI into Tesla, or I'm leaving." They say no. He leaves, slamming the door. Officially: "conflict with Tesla." Unofficially: he's stung.
2019-2022 · The machine takes off. Without Musk, OpenAI opens a tap to investors. Microsoft shows up with a billion-dollar check. Then ten billion. The models come out: GPT-2, GPT-3, DALL·E. And on November 30, 2022, they drop ChatGPT. It's the app that reaches a hundred million people faster than anything in history.
2023 · The revenge begins. Musk creates his own competitor, xAI. He starts attacking OpenAI on Twitter, almost every day. "It's become Microsoft, it's not open at all anymore, they betrayed the mission."
2024 · The first lawsuit. Musk files suit. OpenAI hits back five days later by publishing old Musk emails where he was proposing himself to turn OpenAI into a money machine. Disaster for his narrative. He withdraws the lawsuit in June. Two months later, he comes back with a lawsuit twice as harsh.
2025 · The absurd offer. Musk puts 97.4 billion dollars on the table to buy OpenAI. Altman replies to him in ten words of viral trolling: "no thank you but we will buy twitter for $9.74 billion if you want." Rejected unanimously.
2026 · The trial. Monday of this week, in Oakland. The judge, nine jurors, Musk on the witness stand, Altman waiting his turn. 134 billion at stake. Verdict in late May. And it's right now that I'm telling you all this for real.
Act I · The Menlo Park dinner
To understand this story, you have to understand one thing: it all started with fear. And that fear has a face. Larry Page's, Google's co-founder.
Larry, what have we become, you and me?
2015. Musk is 44. Larry Page and he have been friends for ten years. Not work buddies: real friends. When Musk shuttles between Tesla (Bay Area) and SpaceX (Los Angeles), he sleeps at Larry's place. He has his own room.
At this birthday party, they're talking artificial intelligence in the kitchen, drink in hand. Musk says: "We're going to get burned. If a machine becomes smarter than us, it won't cut us any slack. We need guardrails, rules." Larry shrugs. "Why? If a machine becomes more intelligent, well, good for it. That's evolution." Musk pushes back, gets annoyed. Larry laughs and lets slip: "You're a speciesist." A machine racist, basically.
Eleven years later, in the Oakland courtroom, Musk will say under oath: "The real reason OpenAI exists is because Larry Page called me a speciesist." That night, in the kitchen, a friendship died. And a project started to take root.
An hour of Skype in the dark
Late 2013, Musk learns that Google is going to buy DeepMind, a small British artificial intelligence lab whose lead over everyone else has everybody worried. In a panic, he calls his friend Luke Nosek. That evening, they're both at a party in Los Angeles, in a big villa. Crowds, noise.
They find a broom closet upstairs. Shut the door. Fire up their phones. And call Demis Hassabis, the head of DeepMind, on Skype, for an hour. "Don't sell. We'll find you the money. We'll build something together."
Too late. In January 2014, Google pays 650 million and pockets DeepMind. Musk just lost. Not a little: completely. That night, in the closet, he understood that a single company was potentially going to hold the lever of artificial intelligence. He was not going to let that happen.
July 2015 · "What if we built our own Google?"
Sand Hill Road. The most expensive street in the world after Wall Street. That's where every venture fund in Silicon Valley has its offices. In the middle of the street, the Rosewood Hotel. One summer evening, in a private room, seven people sit down around a table.
Sam Altman presides. At the time, he runs Y Combinator, the incubator that saw the birth of Airbnb, Stripe, Dropbox. He has invited Elon Musk, Greg Brockman (who just left Stripe), and a few top-tier artificial intelligence researchers. The menu: salmon. The topic: "What if we built a project that could stand up to Google?"
The dinner ends late. Everyone agrees. A company is going to be created. Not to make money — to protect humanity, nothing less. Brockman will handle recruiting. Musk will fund it. Altman presides.
The gamble is Ilya Sutskever. A Russian-Canadian from Google, a student of one of the fathers of modern artificial intelligence. The researcher everyone is fighting over. Musk offers him 1.9 million dollars a year. To work at a nonprofit. Unusual. Sutskever signs.
On December 11, 2015, in Montreal, they announce to the world the birth of OpenAI. For the PR, they say they've raised a billion dollars in pledges. (In reality, over six years, they'll only receive 133 million of it. But hey, "a billion" lands harder.)
A few months before the dinner, Sam Altman had written Musk a short email that sums up the whole philosophy of the moment:
"If this technology is going to arrive one way or another, it would still be more reassuring if someone other than Google developed it first."
There it is. There's the whole founding promise. Stop Google from having the monopoly. Not "do artificial intelligence for humanity" for all eternity. That's the wrapping. The content is that email. We'll come back to it. Because it's that little sentence that will, nine years later, kill Musk's lawsuit.
Act II · The break
2016, 2017. For two years, it runs. Brockman manages the team day to day. Sutskever leads the researchers. Altman gives his advice but is still busy with his incubator. Musk, for his part, funds, recruits, posts messages about the dangers of artificial intelligence. A bit removed from the daily grind. But the boss.
First grain of sand, June 2017: Andrej Karpathy, one of OpenAI's best researchers (the one running computer vision research), resigns. To go where? To Tesla. To Musk, that is. Inside OpenAI, things start to grate. A partner helping himself on the way out is irritating.
A few months later, fall arrives and with it, the big blowup. OpenAI needs more money. Much more. To train its models, you need enormous server farms, and that costs fortunes. The question lands on the table: "do we turn OpenAI into a normal company that can take on investors?"
Musk answers: "OK. But on the condition that I'm the boss, that I hold the majority of the shares, and that I control the board of directors."
Sutskever, the scientist, gets right away what that means. On September 20, 2017, he writes to Musk and Altman. A short email. A courageous email:
"The structure you're proposing, Elon, gives you a path to end up with total, personal control of an ultra-powerful artificial intelligence. You, alone. […] I don't think you really want that. But that's what it produces."
Internally, at OpenAI, people start talking about a "dictatorship of the machine." Nobody says that to Musk's face. But they write it in emails to each other. Musk replies within the day, at 2:17 p.m., in four sentences:
"Guys, I'm done. This is the last straw. This discussion is over."
Three months pass. January 2018. Musk comes back with another idea. "We're not going to turn OpenAI into a commercial company. Instead: we fold OpenAI into Tesla. Tesla becomes the parent, OpenAI becomes a department." He forwards the others the email where Karpathy lays out in detail how to do it, and adds at the top: "Exactly this. Tesla is the only company that can stand up to Google."
Refused. Again. This time, it's final. Greg Brockman, who runs the company day to day, writes that evening in his diary (a notebook he's kept since he was 15, and that will be entered as evidence at the 2026 trial):
"This is our only chance to get ourselves out of Elon."
On February 20, 2018, Musk resigns. The official statement talks about a "conflict of interest with Tesla." Internally, nobody buys it. And the detail that stings: Musk had promised 1 billion over several years. He paid 45 million. The tap, he shuts it off on his way out. The gesture hurts — without that money, OpenAI can no longer afford the computers it needs.
The real reason, in two sentences
Officially: Tesla conflict. Unofficially (but documented in black and white in the 2024 emails and Brockman's diary): Musk leaves because he was refused control. It's this gap between the version told to the press and the true version that will, six years later, break his lawsuit.
Act III · Seven years to become a beast
With Musk gone, OpenAI has to survive. And then grow. And then, without meaning to, become something enormous. Seven years, five scenes, and the world is no longer the same.
| Scene | When | What happens |
|---|---|---|
| 01 · Microsoft shows up | March 2019 | OpenAI opens a door to investors · Bill Gates and Satya Nadella sign a billion-dollar check |
| 02 · ChatGPT changes the world | Nov 2022 | Release of a product that was supposed to go unnoticed · 100 million users in two months · Musk takes it right in the face |
| 03 · The first volley | Feb 2024 | Musk files suit · OpenAI hits back with its own emails · Musk takes a beating on the narrative |
| 04 · The heavy artillery | Aug 2024 | New lawsuit, much harsher · Microsoft is named too · 134 billion at stake |
| 05 · The absurd offer | Feb 2025 | Musk puts 97 billion on the table · Altman trolls in ten words · the clash becomes public and permanent |
Microsoft shows up (March 2019)
Without Musk's kitty, OpenAI has a problem. To train its models, you need gigantic server farms, and that costs fortunes. Brockman and Altman do the math: it takes billions a year. A simple nonprofit can't raise that.
So they invent a legal workaround. They create a subsidiary that can, itself, take in money from investors. But with a rule: if you put in ten million, you can never get back more than a billion, ever. Everything else goes back to the nonprofit. It's a weird, hybrid compromise, but it lets them honor the founding promise while attracting capital.
And that's when Microsoft arrives. Bill Gates calls Satya Nadella. The CEO takes a plane to San Francisco. He signs a one-billion-dollar check in July 2019. In exchange, OpenAI will run on Microsoft servers. And Microsoft will get privileged access to the models. Deal done.
Musk watches this from Twitter. He posts in September 2020: "Well, this is the exact opposite of open. OpenAI let itself be bought by Microsoft." His tweets go unnoticed at the time. Nobody cares. ChatGPT doesn't exist yet.
November 30, 2022 · ChatGPT changes the world
The story goes that OpenAI released ChatGPT almost by accident. At headquarters, they thought it would be a tool for tinkerers, that there'd be ten thousand curious people, twenty thousand at most, and that nobody would talk about it after a week.
After five days: a million users. After two months: a hundred million. The product that reaches a hundred million people the fastest in history. Faster than Instagram, than TikTok, than anything. The whole world starts talking to a robot. Your parents get into it. Your teachers too.
Microsoft, right on cue, breaks out the checkbook again. Ten billion, in January 2023. OpenAI's valuation climbs from 29 billion to 157 billion in eighteen months. A rocket.
And Musk, who has just bought Twitter for 44 billion (and renamed it X), watches all this on his phone. Three days after ChatGPT's release, he posts:
"OpenAI started open source and nonprofit. Today, neither one nor the other. By the way, I just learned they were using Twitter's data to train their models. I've cut off the tap."
Cutting off the tap to a former partner, just because you're jealous. That's what it tells you about the Musk of that era. Petty, small-minded, angry. In February 2023, he posts another tweet that will become the slogan of his whole future attack: "I created OpenAI to be a counterweight to Google. Today: closed code, max profit, Microsoft at the controls. The exact opposite of what I wanted."
Five months later, July 2023, he founds his own artificial intelligence company, which he calls xAI. The stated mission: "understand the true nature of the universe." The real mission: "crush OpenAI."
February 29, 2024 · Musk files suit
Musk opens hostilities. He sues in San Francisco against OpenAI, Altman and Brockman. The grievance: "You promised me we'd do artificial intelligence for humanity, for free, open source. Instead, you built a billion-dollar machine locked down by Microsoft. You betrayed me. And along with me, you betrayed my 44 million dollars in donations."
Five days later, OpenAI hits back. And here's where it gets brutal. They publish on their official blog a long article, signed by Altman, Brockman, Sutskever and two others, plainly titled: "OpenAI and Elon Musk." Inside: about a dozen private Musk emails, dated 2015 to 2018, sent to OpenAI back when he was still on the board. And which say, in black and white:
- In 2017, Musk: "Give me the majority, control of the board, and the CEO job."
- In December 2018, Musk: "We need billions a year, immediately, otherwise we give up."
- In February 2018, Musk forwards the idea of folding OpenAI into Tesla, commenting: "Exactly."
In other words: Musk faults OpenAI for doing exactly what he himself was proposing at the time. The conclusion of OpenAI's blog post is elegant and devastating:
"We're sad it's come to this with someone we deeply admired. Someone who pushed us to aim higher, then told us we were going to fail, launched a competitor, and now attacks us because we're moving toward the mission… without him."
The tech press, which until then had been mostly on Musk's side, flips in 48 hours. On June 11, 2024, the day before the hearing where OpenAI was set to ask for the lawsuit to be dismissed, Musk withdraws his suit. Without explanation. Everyone understands he lost the first battle — the one over the narrative.
August 5, 2024 · The heavy artillery
Two months later, Musk comes back. This time in federal court, in Oakland. Harsher, more complete. 26 counts, including one based on the U.S. anti-mafia law. The war machine is launched. The case lands with Judge Yvonne Gonzalez Rogers, who made a name for herself a few years earlier by making Apple sweat in its trial against Epic Games.
In November 2024, he beefs up his case. He adds people on both sides:
- On the plaintiffs' side: his own company xAI, and Shivon Zilis (an executive, former OpenAI board member, and, incidentally, mother of three of his children).
- On the defendants' side: Microsoft, Reid Hoffman (co-founder of LinkedIn) and Dee Templeton (a Microsoft vice president).
On March 4, 2025, Judge Gonzalez Rogers refuses to freeze OpenAI's conversion into a commercial company (Musk had asked for an emergency freeze). But she grants a rare decision: an expedited trial. Plainly put: we're not going to wait four years. We'll do it in fourteen months. The clock starts.
On April 10, 2025, OpenAI officially counterattacks in court. They accuse Musk of running "an unlawful campaign of harassment, interference and disinformation." They ask the judge to order Musk to shut up. Good luck.
February 10, 2025 · The absurd offer
Musk, who loves catching everyone off guard, plays his craziest card. He puts 97.4 billion dollars in cash on OpenAI's table. He offers to buy the nonprofit. For real. In the consortium with him, to carry the check: his usual investor buddies, plus Joe Lonsdale (founder of the fund 8VC) and Ari Emanuel (the boss of Endeavor in Hollywood).
The logic: if Musk buys the nonprofit, he automatically buys the board of directors. So he takes control. So he blocks everything. Brilliant. Nasty.
Four hours later, Altman replies on Twitter. Ten words that instantly become iconic:
"no thank you but we will buy twitter for $9.74 billion if you want"
(Translation: "no thank you, but we'll buy Twitter for 9.74 billion if you want.") That's ten times cheaper. And as a reminder, since Musk bought Twitter for 44 billion in 2022, the platform has lost three-quarters of its value. The blow is unanswerable. Twitter explodes with likes.
Musk replies in a single word. "Swindler." Followed by a doctored video of Altman with an overlay reading: "Scam Altman." The nickname sticks. Even today, Musk calls him that on Twitter. Every day.
On February 14, 2025, OpenAI's board votes unanimously against the offer. Bret Taylor, the board chair (and here's a piquant coincidence: former chair of Twitter's board at the moment Musk bought the company in 2022), issues a very diplomatic statement: "Any reorganization of OpenAI will strengthen our nonprofit and its mission."
The next day, Altman gives an interview to Bloomberg. Question: "What do you think of Elon?" Answer:
"I think he's someone who has operated his whole life from a position of insecurity. I feel for him. I don't think he's a happy person."
The first time Altman, the calm and composed man, attacks Musk head-on. Not on the work. On the psychology. Below the belt. But well landed.
Act IV · The war of tweets
The story of Musk and Altman is also a story that gets told in public one-liners. Three lines in the morning, two sentences at night, and the whole press makes it front-page news. Here are five moments where the words hit hard. I'm giving them to you in the original language (with my translation right below) so you really feel the tone of each. Because they don't talk the same way. Not at all.
"OpenAI was created as an open source (which is why I named it ‘Open’ AI), non-profit company to serve as a counterweight to Google, but now it has become a closed source, maximum-profit company effectively controlled by Microsoft. Not what I intended at all."
"I created OpenAI as an open source company (hence the ‘Open’ in the name), nonprofit, to be a counterweight to Google. Today it's the opposite: closed code, maximum profit, and it's Microsoft holding the controls. This is absolutely not what I wanted."
"I'm still confused as to how a non-profit to which I donated ~$100M somehow became a $30B market cap for-profit. If this is legal, why doesn't everyone do it?"
"I still haven't figured out how a nonprofit I donated about 100 million to became a company worth 30 billion. If it's legal, why doesn't everyone do it?"
"We're sad that it's come to this with someone whom we've deeply admired—someone who inspired us to aim higher, then told us we would fail, started a competitor, and then sued us when we started making meaningful progress towards OpenAI's mission without him."
"We're sad it's come to this with someone we've deeply admired. Someone who inspired us to aim higher, then told us we'd fail, then launched a competitor, and now sues us because we're moving forward on OpenAI's mission… without him."
"no thank you but we will buy twitter for $9.74 billion if you want"
"no thank you, but we'll buy Twitter for 9.74 billion if you want."
"Scam Altman and Greg Stockman stole a charity. Full stop. […] The fundamental question is simply this: Do you want to set legal precedent in the United States that it is ok to loot a charity? If so, you undermine all charitable giving in the United States forever."
"Scam Altman and Greg Stockman stole a charity. Period. […] The real question is this: do we want the U.S. justice system to rule that you can loot a charity? Because if so, the whole system of donations in the US collapses."
See the difference? Musk hits sharp, short, jabs, curse words ("thief," "swindler"), capital letters to shout. Altman writes without capitals (his trademark), gentle, almost tender, but he's a boxer. When he trolls (the 9.74 billion move), it's lethal. When he analyzes Musk, it's psychoanalysis. He goes after your soul. Musk goes after your wallet. Two kinds of blows. Two kinds of men.
Act V · The courtroom
Monday morning, April 27, 2026. Courtroom number 1 of the federal courthouse in Oakland, across from San Francisco. Judge Yvonne Gonzalez Rogers takes her seat. She's 60, short hair, a gaze that doesn't mess around. She already made Apple sweat in 2021 in another huge case (Epic Games vs Apple). Today, she takes on the most-watched trial of the year.
Before we got here, it took two years of legal fighting. And Musk lost a lot. A lot, a lot. Of the twenty-six charges he started with, two were left when the trial opened Monday. Two. Misappropriation of charitable funds and unjust enrichment. The rest, the judge swept away.
But those two are enough. With a court expert who assessed the harm at 134 billion dollars, we're in serious territory. If Musk wins, OpenAI has to cough up that amount to the nonprofit. And he's also asking for Altman's outright removal from the board of directors.
Monday: nine jurors were selected. Tuesday morning: opening statements, the lawyers for both camps lay out their narrative. Musk's lawyer planted his flag in four words: "They stole a charity." OpenAI's lawyer countered: "Musk just wants the control he was denied."
First witness called to the stand: Elon Musk. Himself. They don't waste time. He was on for eight hours over two days. He talked about Larry Page, the 2015 dinner, the break, and why he founded OpenAI. The room listened to him in silence. Except that at every break, he'd grab his phone and tweet during the hearing. Tuesday evening, this post:
"Scam Altman and Greg Stockman stole a charity. Full stop. […] The real question is: do we want to establish in the United States that you can loot a charity? If so, we kill the whole system of donations for good."
Wednesday morning, Judge Gonzalez Rogers showed up irritated. She reprimanded Musk publicly, and threatened him with a gag order — a ban on talking about the trial outside the courtroom. Musk listened. He smiled. He promised nothing.
On the witness benches yet to come: Sam Altman, his former partner. Greg Brockman, whose 2017 diary was entered as evidence. Satya Nadella, Microsoft's CEO. Ilya Sutskever, OpenAI's former chief scientist. And Shivon Zilis, who is a plaintiff with Musk but also the mother of three of his children.
As I'm typing these lines
It's Thursday, April 30. The trial is in its second day of testimony. Musk finished testifying this morning. Altman is due to take the stand next week. Deliberations are expected around May 12. The verdict will land in late May. If you want me to send you the hot take when it comes, subscribe to my newsletter at the bottom of the article.
And you, in all this?
If you're still reading, I can ask you the question at this point: "Why should I care about two men fighting each other in court?" Because it's billionaire drama, yes. But it's also a trial that's going to decide five concrete things for your daily life. Not tomorrow. In the two to five years to come.
1 · Whether a nonprofit can become a money machine
OpenAI started like the Red Cross. It ends like Apple. If the judge says yes to Musk, that means a nonprofit can't do that. And all the other tech nonprofits (Mozilla, which makes Firefox, Wikipedia, the nonprofit that hosts Linux) are watching the verdict the way you watch a disaster movie. If they can one day flip into a commercial company, they can grow much faster. If the judge says no, they stay nonprofits forever, with the constraints that come with it.
2 · Who decides when we've reached "true AI"
In the secret contract between OpenAI and Microsoft, there was a strange clause. If OpenAI declared it had created a "complete" artificial intelligence (one that can do everything a human does), Microsoft lost access to the models. OpenAI alone decided when. The trial forced a change: since April 2026, it's no longer OpenAI that decides. It's a panel of independent experts. A little legal line that changes everything: the power to say "we've made it" moves out of a company's hands and into a committee's. It's a huge precedent for future artificial intelligence laws.
3 · The dream of an "AI for humanity" is dead
The 2015 idea (doing artificial intelligence in a nonprofit, so it doesn't end up in the hands of a single company) doesn't work. Quite simply because it costs too much. A billion dollars per model. Not per year: per model. Whatever the verdict, the message is through: powerful artificial intelligence will be commercial companies. And the only real check will be the law. So politics. So the vote.
4 · The Musk method will inspire copycats
Musk is pioneering a clever tactic. "I attack my competitor over the legal form of its company, while I build my own competitor on the side." If it works, plenty of others will do it. Not just in AI. In pharma, in cars, in defense. The challengers coming up will attack the leaders over their structure. It's going to become a standard weapon.
5 · Five companies to govern the world
Today, the people who can afford the machines to do powerful artificial intelligence are five companies in the world. OpenAI (worth 852 billion). Anthropic (380 billion). Google. xAI (230 billion). Meta. That's it. Five companies decide what machines will be able to do in ten years. The Musk vs OpenAI trial may redistribute a little money. But it changes nothing about that concentration. It's like having five pharaohs. And you, the citizen, you're not in the room.
My honest bet after two days of reading
On the courtroom side: Musk is going to lose. The 2017 emails his former partners published are open-and-shut. In them he was asking himself for OpenAI to become a money machine, and he was writing it to four people cc'd. The likely verdict: symbolic compensation, no dismantling, case closed. On the moral side: he scores a point. OpenAI did betray what it promised. The numbers scream it (852 billion in value, an IPO in preparation, the word "humanity" replaced by the word "mission" in the PR). Musk loses his personal battle. But he put back on the table a question we didn't want to ask ourselves. And that's useful.
If you want to go further
If the story hooked you and you want to dig deeper, here are eight resources that fed me (most in English, but readable with a bit of patience):
- OpenAI · OpenAI and Elon Musk — the official blog post of March 5, 2024, with the founding emails published online. Required reading to understand OpenAI's version.
- Semafor · The secret history of Elon Musk, Sam Altman, and OpenAI — the original scoop (Reed Albergotti, March 24, 2023) that established the "Musk wanted to take control, we refused, he left out of spite" version.
- Wikipedia · Musk v. Altman — the most up-to-date legal timeline, with links to the court documents.
- LessWrong · OpenAI Email Archives — a consolidated archive of all the 2015-2018 emails published via the lawsuit and the OpenAI blog. Raw source, to read like a thriller.
- MIT Technology Review · The messy, secretive reality behind OpenAI's bid to save the world — Karen Hao's big investigation (February 2020) that documented the nonprofit's drift before ChatGPT.
- Lex Fridman Podcast #419 with Sam Altman — ~2h, March 2024, to hear Altman tell the board saga, Musk, his vision. Calm tone.
- Local News Matters · Musk v. Altman trial coverage — day-by-day trial coverage, more precise than the national media.
- Walter Isaacson, Elon Musk (Simon & Schuster, 2023) — the "seen from Musk" version, authorized, useful for the human anecdotes (read with a critical eye, Isaacson was criticized for his lack of distance).
There it is. I've told you this story as simply as I could. The trial continues until late May, and a lot is going to happen between now and the verdict. When the verdict lands, I'll send you my hot take in my newsletter AI Playbook (Friday morning, 9 a.m., once a week, unsubscribe in 1 click). If you don't want to miss it, it's at the bottom of the article. And if you're a lawyer, a journalist, or a direct witness to this story, and you see an error or an angle I'm missing, reply to my email. I read everything, I fix it.
To stay on the same wavelength, also read the letter to my son about the world of AI (the human side), the breakdown of China's 2026-2030 plan (the geopolitical side), or the AI Wars podcast series (USA vs China told in three audio episodes).
FAQ Musk vs OpenAI.
Why is Musk suing OpenAI and Altman?
Musk claims OpenAI betrayed its founding mission. The company was created in 2015 as a nonprofit, with a promise to develop general AI "for the benefit of humanity." In 2019, OpenAI created a subsidiary able to generate a profit (capped-profit), then in October 2025 flipped to a mission-driven company (PBC). Microsoft poured in 13 billion. Musk argues that the ~45 million he gave was for a charity, not to enrich Altman, Brockman and Microsoft.
How much is Musk asking for?
Up to 134 billion dollars in damages, according to the financial expert approved by Judge Yvonne Gonzalez Rogers. A twist: in April 2026 Musk specified that he wants this amount paid to OpenAI's nonprofit foundation, not to himself. He is also demanding that Altman be removed from the board and Brockman from his executive role.
When did the friendship begin, when did it break?
They teamed up in 2015 over a dinner at the Rosewood Sand Hill (Menlo Park) to found OpenAI. The break dates to February 2018: Musk wants to take control, the board (Altman, Brockman, Sutskever) refuses, so Musk proposes merging OpenAI into Tesla — refused too. He leaves saying publicly "conflict of interest with Tesla," but the internal emails published in 2024 show the real reason was the refusal to give him control.
Why did Musk withdraw his 1st lawsuit in June 2024?
Musk filed his first lawsuit on February 29, 2024. OpenAI hit back on March 5, 2024, by publishing the founding emails — which showed Musk himself asking in 2017 for control of a for-profit OpenAI, and writing that the mission required billions in private capital. The narrative reversal was massive. Musk withdrew his lawsuit the day before the hearing where OpenAI was set to ask for dismissal, then refiled two months later in federal court, in a nuclear version (RICO, fraud, 26 counts).
The $97.4 billion offer of February 2025?
On February 10, 2025, a consortium led by Musk (xAI, Baron Capital, Valor, Atreides, Vy Fund, Emanuel Capital, 8VC) filed an unsolicited all-cash offer of 97.4 billion to buy control of the OpenAI nonprofit foundation. The logic: if Musk controls the nonprofit, he controls the board, he blocks the switch to for-profit. The board voted to reject it unanimously on February 14. Altman tweeted: "no thank you but we will buy twitter for $9.74 billion if you want." Musk replied in one word: "Swindler."
Is OpenAI nonprofit or for-profit today?
Since October 28, 2025, a hybrid structure. The nonprofit foundation (OpenAI Foundation) legally controls a mission-driven company (OpenAI Group PBC) that houses the commercial activities. The foundation holds ~26% of the capital, Microsoft ~27%. The historical cap on profits (100x multiple) has been removed. This restructuring is setting up OpenAI's IPO, planned for 2027.
How much is OpenAI worth today?
852 billion dollars, according to the funding round closed on March 31, 2026 (a 122 billion round). For comparison, in early 2023 OpenAI was worth 29 billion. The company makes ~24 billion in annualized revenue (April 2026), with a burn rate of ~17 billion. xAI, the competitor Musk founded in 2023, is valued at 230 billion (before the announced merger with SpaceX that creates a 1,250 billion entity).
Who is the judge?
Yvonne Gonzalez Rogers, a federal judge at the Northern District of California (Oakland). Case number: 4:24-cv-04722-YGR. She is known for presiding over Epic Games vs Apple in 2021. She denied Musk's preliminary injunction in March 2025, but granted an expedited trial. The trial started on April 27, 2026, expected to run ~4 weeks, with a verdict possible in late May 2026.
Why is Musk attacking when he founded xAI?
Musk launched xAI in July 2023, valued at 230 billion in January 2026. OpenAI's defense highlights the irony: Musk faults OpenAI for exactly the path he's taking himself (for-profit, massive private capital, product integration into X and Tesla). Musk's argument at trial: if OpenAI keeps its tax-exempt structure, xAI is fighting on an uneven playing field — what his lawyers call an "unfair competitive disadvantage."
Why does it matter to someone outside tech?
Beyond the billionaire drama, the verdict will settle three things: 1) Whether a nonprofit can transfer its assets to a for-profit (a precedent that touches Mozilla, Wikimedia, every tech foundation). 2) Who controls the definition of AGI — a company board, a panel of experts, a judge? 3) Whether the nonprofit model is viable for developing frontier AI that costs 1 billion per training run. The trajectory of the 5 AI players (OpenAI, Anthropic, Google DeepMind, xAI, Meta) that control 100% of the world's compute depends on the verdict.
Spot a mistake?
Outdated data, a number that moved, an expired source? Write to me at sagnier.jeremy@gmail.com · I fix it within 48h max and I note the update date at the top of the article. Sensitive topic, fast-moving sources — feedback from specialists is worth a thousand times my 2 days of reading. I read everything, I reply.
Shall we keep going?
I test AI for real and I share what works, no jargon, no hype. If this article helped you, the simplest way to not miss anything is my Friday letter. And if you have a question or a doubt: reply to me, I read everything.